US20060005105A1 - Decoder and decoding method for decoding low-density parity-check codes with parity check matrix - Google Patents
Decoder and decoding method for decoding low-density parity-check codes with parity check matrix Download PDFInfo
- Publication number
- US20060005105A1 US20060005105A1 US11/168,329 US16832905A US2006005105A1 US 20060005105 A1 US20060005105 A1 US 20060005105A1 US 16832905 A US16832905 A US 16832905A US 2006005105 A1 US2006005105 A1 US 2006005105A1
- Authority
- US
- United States
- Prior art keywords
- parity
- bit
- check matrix
- column
- positions
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
- H03M13/13—Linear codes
- H03M13/19—Single error correction without using particular properties of the cyclic codes, e.g. Hamming codes, extended or generalised Hamming codes
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
- H03M13/11—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
- H03M13/1102—Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
- H03M13/1148—Structural properties of the code parity-check or generator matrix
- H03M13/116—Quasi-cyclic LDPC [QC-LDPC] codes, i.e. the parity-check matrix being composed of permutation or circulant sub-matrices
- H03M13/1168—Quasi-cyclic LDPC [QC-LDPC] codes, i.e. the parity-check matrix being composed of permutation or circulant sub-matrices wherein the sub-matrices have column and row weights greater than one, e.g. multi-diagonal sub-matrices
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
- H03M13/11—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
- H03M13/11—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
- H03M13/1102—Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
- H03M13/1105—Decoding
- H03M13/1131—Scheduling of bit node or check node processing
- H03M13/1137—Partly parallel processing, i.e. sub-blocks or sub-groups of nodes being processed in parallel
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
- H03M13/11—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
- H03M13/1102—Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
- H03M13/1105—Decoding
- H03M13/1131—Scheduling of bit node or check node processing
- H03M13/114—Shuffled, staggered, layered or turbo decoding schedules
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
- H03M13/11—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
- H03M13/1102—Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
- H03M13/1148—Structural properties of the code parity-check or generator matrix
- H03M13/116—Quasi-cyclic LDPC [QC-LDPC] codes, i.e. the parity-check matrix being composed of permutation or circulant sub-matrices
Definitions
- the present invention relates to a decoder and decoding method for decoding low-density parity-check codes with parity check matrices.
- LDPC codes low-density parity-check codes
- Message-passing algorithms are known as a technique for decoding LDPC codes from, for example, Andrew J. Blanksby and Chris J. Howland, “A 690-mW 1-Gb/s 1024-b, Rate-1 ⁇ 2 Low-Density Parity-Check Code Decoder”, IEEE Journal of Solid-State Circuits, Vol. 37, No. 3, pp. 404-412, March 2002.
- bit update computation which is performed for each row of a check matrix (parity-check matrix)
- parity update computation which is performed for each column of the matrix
- bit update computation and parity update computation depend upon each other. Therefore, in the prior art, parity update computation of one loop cannot be started unless bit update computation of one loop is finished, and bit update computation of the next loop cannot be started unless parity update computation of the one loop is finished.
- a low-density parity-check code decoder comprises (r ⁇ m) bit-processing units, m parity-processing units and a controller.
- the (r ⁇ m) bit-processing units are configured to perform bit update computation for sequentially updating bit information corresponding to column positions included in the respective rows of a parity-check matrix, the parity-check matrix being formed of (r ⁇ s) permuted matrices each having an array of (m ⁇ m) and being divided into s column blocks, a bit at each of the column positions being set to “1”.
- the m parity-processing units are configured to perform parity update computation for updating parity information corresponding to row positions in m columns of each column block of the parity-check matrix, a bit at each of the row positions being set to “1”.
- the controller is configured to cause, whenever the (r ⁇ m) bit-processing units have finished bit update computation for m column positions in the respective rows of the parity-check matrix, the m parity-processing units to perform parity update computation corresponding to m columns of one of the s column blocks to which the m column positions belong. Further, the controller is configured to cause, after the parity-processing units finish parity update computation for m columns of a first one of the s column blocks, the (r ⁇ m) bit-processing units to start next bit update computation.
- FIG. 2 is a block diagram illustrating the configuration of an LDPC-code decoder according to the embodiment
- FIG. 5 is a view illustrating information input and output in each cycle during bit update computation by the bit-processing units 10 - 1 to 10 -M and parity update computation by the parity-processing units 20 - 1 to 20 - m;
- FIG. 6 is a view illustrating changes in the states of the areas of a memory 30 corresponding to respective (3 ⁇ 5 (r ⁇ s)) permuted matrices that provide a check matrix H;
- FIG. 7 is a view useful in explaining a schedule for decoding an LDPC code employed in the embodiment.
- Each permuted matrix I g,h is a sub-matrix of the check matrix H.
- Each permuted matrix I g,h is acquired by iterating, a number of times, exchange of two different rows or columns included in unit matrices I each having an array of (m ⁇ m).
- the number M of rows of the check matrix H is rm
- the number N of columns of the matrix H is sm.
- the check matrix H is divided into s column blocks (i.e., the 1 st to the s th column blocks).
- Permuted matrices I 1,h , I 2,h and I 3,h belong to the h th column block.
- each row of the check matrix H can be regarded as a parity-check equation.
- the parity-check equation corresponding to the i th row of the check matrix H is called an i th parity-check equation.
- the N-bit LDPC code must satisfy all parity-check equations of the check matrix H, i.e., the first to the M th parity-check equations.
- each permutated matrix I g,h indicate the locus of the positions of “1” bits.
- the numbers of “1” bits contained in each row and each column of the check matrix H are s and r, respectively. r and s are set to satisfy r ⁇ s. Accordingly, the number of “1” bits in each row is greater than that in each column.
- r i,j is acquired by passing corresponding parity information (reliability information) q i,j to each bit of the i th row of the check matrix H (except for the j th bit in the i th row).
- r i,j is bit information (reliability information) indicating reliability concerning code bit x j corresponding to the j th bit in the i th row.
- r i,j does not indicate the probability itself that code bit x j is 0 or 1, but the logarithmic value of the probability. The reason why such a logarithmic value is used is that it enables multiplication/division to be replaced with addition/subtraction.
- M(i) represents a set of column positions of “1” bits in the i th row of the check matrix H.
- M(i) ⁇ j represents M(i) except for j.
- j′ ⁇ M(i) ⁇ j represents an arbitrary column position included in M(i) except for j.
- q i,j′ is parity information (reliability information) indicating reliability concerning code bit x j , corresponding to the j th bit in the i th row of the check matrix H. Assume that q i,j′ does not indicate the probability itself that code bit x j′ is 0 or 1, but the logarithmic value of the probability.
- sgn (q i,j′ ) represents the sign of q i,j′ .
- ⁇ ( ) represents a function (hereinafter referred to as “ ⁇ function”) for probability computation.
- q i,j is acquired by passing, from a certain bit node to a corresponding parity node, bit information (reliability information) r i′,j corresponding to each bit of the j th column of the check matrix H (except for the i th bit in the j th column).
- q i,j is parity information (reliability information) indicating reliability concerning code bit x j corresponding to the j th bit in the i th row of the check matrix H.
- q i,j is used as the above-mentioned q i,j′ for bit update computation expressed by the equation (1).
- p j represents the logarithmic value of the probability (initial probability) that the j th bit x j is 1 or 0, which is estimated from an LDPC code when it is output through a communication channel.
- the initial value of q i,j corresponding to each “1” bit in the check matrix H is identical to p j .
- N(j) represents a set of row positions of “1” bits in the j th column-of the check matrix H.
- N(j) ⁇ i represents N(j) except for i. i′ ⁇ N(j) ⁇ i represents an arbitrary row position included in N(j) except for i.
- r i′,j is bit information (reliability information) indicating reliability concerning the j th bit x j in the i th row of the check matrix H.
- r i,j updated by computation using the equation (1) is used as r i′,j .
- FIG. 2 is a block diagram illustrating the configuration of an LDPC-code decoder according to the embodiment.
- the bit-processing units 10 - 1 to 10 -M are provided for the first to M th rows of the check matrix H, respectively.
- each parity-processing unit 20 - k is used to perform parity update computation on the k th column, (m+k) th column, (2m+k) th column, (3m +k) th column and (4m +k) th column of the check matrix H.
- the parity-processing unit 20 - 1 is used to perform parity update computation on the first column, (m+ 1 ) th column, (2m+1) th column, (3m+1) th column and (4m+1) th column of the check matrix H.
- the parity-processing unit 20 - m is used to perform parity update computation on the m th column, 2m th column, 3m th column, 4m th column and 5m th column (the N th column) of the check matrix H.
- the memory 30 is used to store bit information and parity information.
- the bit information is calculated by the bit-processing units 10 - 1 to 10 -M.
- the parity information is calculated by the parity-processing units 20 - 1 to 20 - m .
- the (3 ⁇ 5) memory units can be simultaneously accessed.
- the state of the storage area of the memory 30 corresponding to each of permuted matrices I g,h is represented by Q g,h or R g,h .
- Q g,h indicates that parity information is stored in a storage area of the memory 30 corresponding to Q g,h .
- R g,h indicates that bit information is stored in a storage area of the memory 30 corresponding to R g,h .
- the controller 40 controls decoding processing of an LDPC code input to the LDPC-code decoder. Specifically, the controller 40 functions as a sequencer for controlling the operations of the bit-processing units 10 - 1 to 10 -M and parity-processing unit 20 - 1 to 20 - m.
- Bit update computation expressed by the equation (1) mainly comprises a product part of H operation, a ⁇ function part and a summation part of ⁇ operation.
- the ⁇ function part i.e., the ⁇ ( ) operation, can be realized using a lookup table.
- the bit-processing unit 10 - i is formed of a summation computation circuit 110 , lookup table (LUT) 120 and sign computation circuit 130 .
- the summation computation circuit 110 corresponds to the summation part.
- the lookup table 120 and sign computation circuit 130 correspond to the ⁇ function part and product part, respectively.
- the summation part can be divided into an adder section and subtracter section.
- the adder section calculates the sum of ⁇ (q i,j ) values corresponding to positions j in the i th row of the check matrix H, at which the bit is set to “1”.
- the subtracter section subtracts, from the sum calculated by the adder section, ⁇ (q i,j ) corresponding to position j in the i th row at which the bit is set to “1”.
- the summation computation circuit 110 is formed of an addition circuit 111 and subtraction circuit 112 .
- the addition circuit 111 comprises a pair of registers 111 a , multiplexers (MUX) 111 b and 111 b , and adder 111 d .
- the registers 111 a are used to hold the intermediate result (or final result) of addition processing for calculating the sum of ⁇ (q i,j ) values.
- the final result of addition processing is the last intermediate result, and indicates the sum of ⁇ (q i,j ) values.
- the other register 111 a is used to hold a new intermediate result of addition processing used for calculating a new sum of ⁇ (q i,j ) values.
- the functions of the registers 111 a as respective registers for holding the sum and the intermediate result of ⁇ (q i,j ) values are switched each time the sum of ⁇ (q i,j ) values is calculated.
- the multiplexer 111 b sequentially selects ⁇ (q i,j ) corresponding to each position (column position) j in the i th row of the check matrix H, at which the bit is set to “1”. Position j in the i th row of the check matrix H, at which the bit is set to “1”, can be determined uniquely in units of rows. Accordingly, ⁇ (q i,j ) corresponding to position j in the i th row at which the bit is set to “1” can be determined uniquely.
- the multiplexer 111 c selects the intermediate result (the result of the preceding addition process) held by one of the registers 111 a .
- the adder 111 d adds (q i,j ) selected by the multiplexer 111 b to the intermediate result selected by the multiplexer 111 c .
- the addition result of the adder 111 d indicates a new intermediate result of addition processing for calculating the sum of ⁇ (q i,j ) values.
- the intermediate result held by the register 111 a and used for addition of the adder 111 d is updated to the addition result of the adder 111 d , i.e., the new intermediate result of addition processing for calculating the sum of ⁇ (q i,j ) values.
- the multiplexer 111 b sequentially selects ⁇ (q i,j ) corresponding to each position j in the i th row of the check matrix H, at which the bit is set to “1”. Further, each time the multiplexer 111 b selects ⁇ (q i,j ), the addition circuit 111 adds the selected ⁇ (q i,j ) to the intermediate result held by one of the registers 111 a (i.e., the intermediate result selected by the multiplexer 111 c ). Each time this addition is performed, the intermediate result of the register 111 a used for the addition is updated to a new intermediate result as the addition result.
- processing of one row of each permuted matrices I g,h in the addition circuit 111 can be performed by one cycle.
- one cycle is time Tcyc
- the subtraction circuit 112 comprises multiplexers (MUX) 112 a and 112 b , and a subtracter 112 c .
- the multiplexer 112 a sequentially selects ⁇ (q i,j ) corresponding to each position j in the i th row of the check matrix H, at which the bit is set to “1”, in cycles following the cycle in which the sum of ⁇ (q i,j ) values corresponding to positions j in the i th row of the check matrix H, at which the bit is set to “1”, is held by one of the registers 111 a .
- the multiplexer 112 b selects the output of the register 111 a , i.e., the above-mentioned sum.
- the subtracter 112 c subtracts ⁇ (q i,j ) selected by the multiplexer 112 a from the sum selected by the multiplexer 112 b .
- the sum of ⁇ (q i,j ) values corresponding to all positions j′ that satisfy M(i) ⁇ j can be acquired.
- the subtraction, by the subtracter 112 c , of ⁇ (q i,j ) from the sum held by the one register 111 a is performed concerning ⁇ (q i,j ) values corresponding to all positions j in the i th row of the check matrix H, at which the bit is set to “1”.
- the subtraction circuit 112 sequentially calculates the sum of ⁇ (q i,j′ ) values corresponding to all positions j′ that satisfy M(i) ⁇ j.
- the addition circuit 111 can perform the next addition process.
- the multiplexer 111 c selects the other of the registers 111 a (i.e., the register that is not selected by the multiplexer 112 b during the subtraction process by the subtraction circuit 112 ).
- the content of the other register is updated to the addition result of the adder 111 d included in the addition circuit 111 .
- the next addition process by the addition circuit 111 is started after the parity-processing unit 20 - k acquires new ⁇ (q i,j ) corresponding to each new position i in the first to m th rows of the check matrix H, at which the bit is set to “1”.
- the lookup table 120 is used to convert the subtraction result of the subtracter 112 (i.e., the computation result of the sum computation circuit 110 ) into an operation value of the ⁇ function, i.e., ⁇ ′ (subtraction result).
- the lookup table 120 is referred to, using the subtraction result of the subtracter 112 .
- the subtraction result is x.
- the entry of the lookup table 120 designated by x prestores ⁇ (x) unique to x. Accordingly, ⁇ (x) can be acquired from the lookup table 120 simply by referring to the table 120 using the subtraction result x of the subtracter 112 .
- the sign computation circuit 130 comprises a pair of registers 131 , multiplexers (MUX) 132 and 133 , exclusive-OR circuit 134 , multiplexers (MUX) 135 and 136 , exclusive-OR circuit 137 and multiplexer (MUX) 138 .
- the registers 131 are each a register of, for example, 1 bit used to hold the intermediate result (or final result) of sign computation. When one of the registers 131 holds the final result of sign computation, the other register is used to hold the intermediate result of new sign computation.
- the multiplexer 132 sequentially selects an MSB of q i,j , i.e., a sign, corresponding to each position j in the i th row of the check matrix H, at which the bit is set to “1”. Each MSB of q i,j will be often expressed by MSB(q i,j ).
- the multiplexer 133 selects the intermediate result (sign) of sign computation so far held by one of the registers 111 a .
- the exclusive-OR circuit 134 acquires the exclusive-OR of the sign selected by the multiplexer 132 and the intermediate result (sign) of sign computation so far.
- the exclusive-OR result of the exclusive-OR circuit 134 indicates a new intermediate result of sign computation.
- the intermediate result in the register 131 used for the operation of the exclusive-OR circuit 134 is updated to the exclusive-OR result of the exclusive-OR circuit 134 , i.e., the new intermediate result of sign computation.
- the multiplexer 132 sequentially selects MSB(q i,j ) corresponding to each position j in the i th row of the check matrix H, at which the bit is set to “1”.
- the sign computation circuit 130 executes the operation of acquiring the exclusive-OR of the selected MSB(q i,j ) and the intermediate result (sign) of sign computation held by the register 131 at this time. Further, each time exclusive-OR is computed, the intermediate result held by the register 131 and used for the exclusive-OR computation is updated to a new intermediate result as the result of the computation.
- the exclusive-OR of MSB(q i,j ) corresponding to the last position j in the i th row and the intermediate result held by the register 131 at this time is computed.
- the result (final result) of this exclusive-OR computation indicates the product of the signs of q i,j corresponding to all positions j in the i th row at which the bit is set to “1”, i.e., the product of the signs of q i,j corresponding to all positions j included in M(i).
- the multiplexer 135 sequentially selects MSB(q i,j ) corresponding to each position j in the i th row of the check matrix H, at which the bit is set to “1”.
- the selection operation of the multiplexer 135 is executed in cycles following the cycle in which the exclusive-OR concerning MSB(q i,j ) values corresponding to all positions j included in M(i) (i.e., the final result of the exclusive-OR computation indicating the product) is held by one of the registers 131 .
- the multiplexer 136 selects the output of the register 131 , i.e., the above-mentioned exclusive-OR (the final result of the exclusive-OR computation).
- the exclusive-OR circuit 137 acquires the exclusive-OR of the exclusive-OR selected by the multiplexer 136 and MSB(q i,j ) selected by the multiplexer 135 . As a result, the product of the signs of q i,j′ values corresponding to all positions j′ that satisfy M(i) ⁇ j can be acquired.
- the operation of the exclusive-OR circuit 137 for acquiring the exclusive-OR of the product (held by the one register 131 ) and MSB(q i,j ) is performed concerning MSB(q i,j ) values corresponding to all positions j in the i th row of the check matrix H, at which the bit is set to “1”.
- the sign computation circuit 130 sequentially computes the product of MSB(q i,j ) values corresponding to all positions j′ that satisfy M(i) ⁇ j.
- the multiplexer 138 selects, as r i,j , the output of the lookup table 120 or a value acquired by inverting the sign of this output.
- the parity-processing unit 20 - k is constructed in light of the feature that the number (3 in the embodiment) of “1” bits in each column of the check matrix H is smaller than the number (5 in the embodiment) of “1” bits in each row of the check matrix H.
- the parity-processing unit 20 - k is configured to simultaneously perform parity update computation on all columns j, using r i,j corresponding to each position (row position) i in the j th column of the check matrix H, at which the bit is set to “1”.
- the parity-processing unit 20 - k comprises three adders 201 , 202 and 203 , three subtracters 204 , 205 and 206 , and three lookup tables (LUT) 207 , 208 and 209 as shown in FIG. 4 .
- the adder 201 adds r i1,j to r i2,j
- the adder 202 adds r i3,j to p j .
- r i1,j , r i2,j and r i3,j represent bit information items corresponding to bit positions i 1 , i 2 and i 3 in the j th column (column j) of the check matrix H, at which the bit is set to “1”.
- the j th column of the check matrix H is positioned in three permuted matrices I 1,h , I 2,h and I 3,h (of the h th column block).
- each of the permuted matrices I 1,h , I 2,h and I 3,h contains one position in the j th column, at which the bit is set to “1”.
- r i1,j , r i2,j and r i3,j represent bit information items corresponding to bit positions in the permuted matrices I 1,h , I 2,h and I 3,h and in the j th column of the check matrix H, at which the bit is set to “1”.
- the bit positions i 1 , i 2 and i 3 in the j th column of the check matrix H, at which the bit is set to “1”, can be uniquely determined in units of rows.
- r i1,j , r i2,j and r i3,j corresponding to the bit positions i 1 , i 2 and i 3 in the j th column, at which the bit is set to “1”, can be uniquely determined.
- the adder 203 adds up the addition results of the adders 201 and 202 . Namely, the adder 203 sums up p j and the total of r i1,j , r i2,j and r i3,j . Thus, the adders 201 to 203 provide an addition circuit for summing up p j and the total of r i1,j , r i2,j and r i3,j .
- the addition result of the adder 203 indicates the probability q j of bit x j in the LDPC code being 1 or 0, which is estimated at the present stage.
- the subtracter 204 subtracts r i1,j from the addition result of the adder 203 .
- the subtraction result of the subtracter 204 indicates updated parity information q i1,j .
- the subtracter 205 subtracts r i2,j from the addition result of the adder 203 .
- the subtraction result of the subtracter 205 indicates updated parity information q i2,j .
- the subtracter 206 subtracts r i3,j from the addition result of the adder 203 .
- the subtraction result of the subtracter 206 indicates updated parity information q i3,j .
- MSB(q i1,j ), MSB(q i2,j ) and MSB(q i3,j ) can be acquired.
- ⁇ (q i1,j ), ⁇ (q i2,j ) and ⁇ (q i3,j ) can also be simultaneously acquired by simultaneously referring to the lookup tables 207 , 208 and 209 using q i1,j , q i2,j and q i3,j as the subtraction results of the subtracters 204 , 205 and 206 .
- ⁇ (q i1,j ), ⁇ (q i2,j ) and ⁇ (q i3,j ) are used as ⁇ (q i,j ) for bit update computation in each bit-processing unit 10 - i.
- one feature of the embodiment lies in that the lookup tables 207 , 208 and 209 are provided at the output side of each parity-processing unit 20 - k .
- respective lookup tables can be provided at the output side of the multiplexers 132 and 135 in each bit-processing unit 10 - i .
- three lookup tables are necessary for each bit-processing unit (i.e., for processing each row), although no lookup table is necessary for each parity-processing unit.
- the number of required lookup tables can be reduced from 9m to 6m.
- FIGS. 5 and 6 a description will be given of the flow of computation performed in the bit-processing units 10 - 1 to 10 -M and parity-processing units 20 - 1 to 20 - m .
- the bit-processing units 10 - 1 to 10 -M and parity-processing units 20 - 1 to 20 - m are controlled by the controller 40 incorporated in the LDPC-code decoder shown in FIG. 2 .
- FIG. 5 shows input/output information in each cycle during bit update computation by the bit-processing units 10 - 1 to 10 -M and parity update computation by the parity-processing units 20 - 1 to 20 - m .
- a set of ⁇ (q i,j ) values and a set of r i,j values are represented by Q g,h and R g,h , respectively.
- Q g,h includes MSB(q i,j ).
- the controller 40 executes an initialization process.
- the controller 40 stores ⁇ (p j ) and sgn(p j ) into each memory unit of the memory 30 .
- p j represents a logarithmic value of probability acquired from the output of the communication channel. Concerning all j values, ⁇ (p j ) is set as the initial value of each of ⁇ (q i1,j ), ⁇ (q i2,j ) and ⁇ (q i3,j ).
- the M bit-processing units 10 - 1 to 10 -M start addition processing. Specifically, MSB(q i,j1 ) and ⁇ (q i,j1 ) are stored into the registers 131 and 111 a of each of the bit-processing units 10 - 1 to 10 -M, respectively.
- the first bit position ji belongs to the first column block.
- the multiplexer 111 b in the addition circuit 111 of each of the bit-processing units 10 - 1 to 10 -M selects ⁇ (q i,j2 ).
- the addition circuit 111 performs addition processing, using ⁇ (q i,j2 ).
- the adder 111 d of the addition circuit 111 adds up the value stored in the register 111 a in the first cycle, and ⁇ (q i,j2 ) selected by the multiplexer 111 b .
- the addition result of the adder 111 d i.e., the sum of the value ( ⁇ (q i,j1 )) stored in the register 111 a and ⁇ (q i,j2 ), is stored into the register 111 a .
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M selects MSB(q i,j2 ), using the multiplexer 132 .
- the exclusive-OR circuit 134 of the sign computation circuit 130 executes exclusive-OR computation.
- the exclusive-OR circuit 134 calculates the exclusive-OR of the value (MSB(q i,j1 )) stored in the register 131 in the first cycle, and MSB(q i,j2 ) selected by the multiplexer 132 .
- the computation result of the exclusive-OR circuit 134 i.e., the exclusive-OR of the value stored in the register 131 and MSB(q i,j2 ), is stored in the register 131 .
- the second bit position j 2 belongs to the second column block.
- the multiplexer 111 b in the addition circuit 111 of each of the bit-processing units 10 - 1 to 10 -M selects ⁇ (q i,j3 ).
- the addition circuit 111 performs addition processing, using ⁇ (q i,j3 ).
- the adder 111 d of the addition circuit 111 adds up the value stored in the register 111 a in the second cycle, and ⁇ (q i,j3 ) selected by the multiplexer 111 b .
- the addition result of the adder 111 d i.e., the sum of the value stored in the register 111 a and ⁇ (q i,j3 ), is stored into the register 111 a .
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M selects MSB(q i,j3 ), using the multiplexer 132 .
- the exclusive-OR circuit 134 of the sign computation circuit 130 executes exclusive-OR computation. Namely, the exclusive-OR circuit 134 calculates the exclusive-OR of the value stored in the register 131 in the second cycle, and MSB(q i,j3 ) selected by the multiplexer 132 .
- the computation result of the exclusive-OR circuit 134 i.e., the exclusive-OR of the value stored in the register 131 and MSB(q i,j3 ), is stored in the register 131 .
- the third bit position j 3 belongs to the third column block.
- the multiplexer 111 b in the addition circuit 111 of each of the bit-processing units 10 - 1 to 10 -M selects ⁇ (q i,j4 ).
- the addition circuit 111 performs addition processing, using ⁇ (q i,j4 ).
- the adder 111 d of the addition circuit 111 adds up the value stored in the register 111 a in the third cycle, and ⁇ (q i,j 4 ) selected by the multiplexer 111 b .
- the addition result of the adder 111 d i.e., the sum of the value stored in the register 111 a and ⁇ (q i,j4 ), is stored into the register 111 a .
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M selects MSB(q i,j4 ), using the multiplexer 132 .
- the exclusive-OR circuit 134 of the sign computation circuit 130 executes exclusive-OR computation. Namely, the exclusive-OR circuit 134 calculates the exclusive-OR of the value stored in the register 131 in the third cycle, and MSB(q i,j4 ) selected by the multiplexer 132 .
- the computation result of the exclusive-OR circuit 134 i.e., the exclusive-OR of the value stored in the register 131 and MSB(q i,j4 ), is stored in the register 131 .
- the fourth bit position j 4 belongs to the fourth column block.
- the multiplexer 111 b in the addition circuit 111 of each of the bit-processing units 10 - 1 to 10 -M selects *(q i,j5 ).
- the addition circuit 111 performs addition processing, using ⁇ (q i,j5 ).
- the adder 111 d of the addition circuit 111 adds up the value stored in the register 111 a in the fourth cycle, and ⁇ (q i,j5 ) selected by the multiplexer 111 b .
- the addition result of the adder 111 d i.e., the sum of the value stored in the register 111 a and ⁇ (q i,j5 ), is stored into the register 111 a .
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M selects MSB(q i,j5 ), using the multiplexer 132 .
- the exclusive-OR circuit 134 of the sign computation circuit 130 executes exclusive-OR computation. Namely, the exclusive-OR circuit 134 calculates the exclusive-OR of the value stored in the register 131 in the fourth cycle, and MSB(q i,j5 ) selected by the multiplexer 132 .
- the computation result of the exclusive-OR circuit 134 i.e., the exclusive-OR of the value stored in the register 131 and MSB(q i,j5 ), is stored in the register 131 .
- the fifth bit position j 5 belongs to the fifth column block.
- the subtraction circuit 112 of each of the bit-processing units 10 - 1 to 10 -M starts a subtraction process.
- the final result means the sum of ⁇ (q i,j ) values corresponding to positions in the i th row of the check matrix H, at which the bit is set to “1”.
- ⁇ (q i,j1 ) is subtracted from the final result of the addition process.
- the multiplexer 112 b of the subtraction circuit 112 selects the one of the registers 111 a that holds the final result of the addition process.
- the multiplexer 112 a of the subtraction circuit 112 selects ⁇ (q i,j1 ).
- the subtracter 112c of the subtraction circuit 112 subtracts ⁇ (q i,j1 ) selected by the multiplexer 112 a from the final result of the addition process selected by the multiplexer 112 b.
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M excludes MSB(q i,j1 ) from the final result of the exclusive-OR computation.
- the multiplexer 136 in the sign computation circuit 130 selects the one of the registers 131 that holds the final result of the exclusive-OR computation.
- the final result of the exclusive-OR computation means the exclusive-ORs of MSB(q i,j ) values corresponding to positions j in the i th row of the check matrix H, at which the bit is set to “1”.
- the multiplexer 135 in the sign computation circuit 130 selects MSB(q i,j1 ).
- the exclusive-OR circuit 137 of the sign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by the multiplexer 136 and MSB(q i,j1 ) selected by the multiplexer 135 .
- the sign of information output from the multiplexer 138 is determined, whereby bit information r i,j is acquired.
- the bit information r i,j is stored in the memory 30 at a position determined from the i th row and j th column corresponding to the bit information r i,j .
- R 1,1 , R 2,1 and R 3,1 are generated and stored in the memory 30 .
- R 1,1 , R 2,1 and R 3,1 represent a set of r i,j values corresponding to the bit positions in the permuted matrices I 1,1 , I 2,1 and I 3,1, which correspond to the 1 st to m th columns (of the first column block) included in the check matrix H, and at which the bit is set to “1”.
- the subtraction circuit 112 of each of the bit-processing units 10 - 1 to 10 -M starts a subtraction process.
- the final result means the sum of ⁇ (q i,j ) values corresponding to positions in the i th row of the check matrix H, at which the bit is set to “1”.
- ⁇ (q i,j2 ) is subtracted from the final result of the addition process.
- the multiplexer 112 b of the subtraction circuit 112 selects the one of the registers 111 a that holds the final result of the addition process.
- the multiplexer 112 a of the subtraction circuit 112 selects ⁇ (q i,j2 ).
- the subtracter 112 c of the subtraction circuit 112 subtracts ⁇ (q i,j2 ) selected by the multiplexer 112 a from the final result of the addition process selected by the multiplexer 112 b.
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M excludes MSB(q i,j2 ) from the final result of the exclusive-OR computation.
- the multiplexer 136 in the sign computation circuit 130 selects the one of the registers 131 that holds the final result of the exclusive-OR computation.
- the multiplexer 135 in the sign computation circuit 130 selects MSB(q i,j2 ).
- the exclusive-OR circuit 137 of the sign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by the multiplexer 136 and MSB(q i,j2 ) selected by the multiplexer 135 .
- bit information r i,j is acquired.
- the bit information r i,j is stored in the memory 30 at a position determined from the i th row and j th column corresponding to the bit information r i,j .
- R 1,2 , R 2,2 and R 3,2 are generated and stored in the memory 30 .
- R 1,2 , R 2,2 and R 3,2 represent a set of r i,j values corresponding to the bit positions in the permuted matrices I 1,2 , I 2,2 and I 3,2 , which correspond to the (m+1) th to 2m th columns (of the second column block) included in the check matrix H, and at which the bit is set to “1”.
- each element r i,j in R 1,2 R 2,2 and R 3,2 stored in the memory 30 can be utilized. Accordingly, in the seventh cycle, the parity-processing units 20 - 1 to 20 - m start parity update computation under the control of the controller 40 , utilizing each element r i,j in R 1,2 , R 2,2 and R 3,2 .
- parity update computation can be started before bit update computation corresponding to all bits in each of the 1 st to M th rows of the check matrix H is finished.
- parity update computation can be started when only bit update computation is finished which corresponds to each bit position in the permuted matrices I 1,1 , I 2,1 and I 3,1 (of the first column block) included in the check matrix H, at which the bit is set to “1”.
- bit update computation and parity update computation can be processed using a pipeline.
- each element q i,j in updated Q′ 1,2 , Q′ 2,2 and Q′ 3,2 can be utilized. Accordingly, in the eighth cycle, the bit-processing units 10 - 1 to 10 -M start next bit update computation under the control of the controller 40 , thereby performing the same addition process as in the first cycle. However, the result of the addition process is held by the register 111 a other than the register 111 a used to hold the result of the addition process of the first cycle (the first to fifth cycles).
- bit update computation can be started before parity update computation corresponding to all bits in each of the 1st to N th columns of the check matrix H is finished.
- next bit update computation can be started when only parity update computation is finished which corresponds to each bit position in the permuted matrices I 1,1 , I 2,1 and I 3,1 (of the first column block) included in the check matrix H (i.e., when parity update computation has been performed concerning the first m columns of the check matrix H).
- the combination of bit update computation and parity update computation can be iterated using a pipeline.
- a subtraction process by the bit-processing units 10 - 1 to 10 -M is executed in parallel with the above addition process.
- the final result of the addition process held by the one of the register 111 a in each of the bit-processing unit 10 - 1 to 10 -M, and Q 1,3 , Q 2,3 and Q 3,3 stored in the memory 30 are used.
- the final result of the addition process means the sum of ⁇ (q i,j ) values corresponding to positions in the i th row of the check matrix H, at which the bit is set to “1”.
- ⁇ (q i,j3 ) is subtracted from the final result of the addition process.
- the multiplexer 112 b of the subtraction circuit 112 selects the one of the registers 111 a that holds the final result of the addition process.
- the multiplexer 112 a of the subtraction circuit 112 selects ⁇ (q i,j3 ).
- the subtracter 112 c of the subtraction circuit 112 subtracts ⁇ (q i,j3 ) selected by the multiplexer 112 a from the final result of the addition process selected by the multiplexer 112 b.
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M excludes MSB(q i,j3 ) from the final result of the exclusive-OR computation.
- the multiplexer 136 in the sign computation circuit 130 selects the one of the registers 131 that holds the final result of the exclusive-OR computation.
- the multiplexer 135 in the sign computation circuit 130 selects MSB(q i,j3 ).
- the exclusive-OR circuit 137 of the sign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by the multiplexer 136 and MSB(q i,j3 ) selected by the multiplexer 135 .
- bit information r i,j is acquired.
- the bit information r i,j is stored in the memory 30 at a position determined from the i th row and j th column corresponding to the bit information r i,j .
- R 1,3 , R 2,3 and R 3,3 are generated and stored in the memory 30 .
- R 1,3 , R 2,3 and R 3,3 represent a set of r i,j values corresponding to the bit positions in the permuted matrices I 1,3 , I 2,3 and I 3,3 , which correspond to the (2m+1) th to 3m th columns (of the third column block) included in the check matrix H, and at which the bit is set to “1”.
- the parity-processing units 20 - 1 to 20 - m perform parity update computation, utilizing each element r i,j in R 1,2 , R 2,2 and R 3,2 stored in the memory 30 in the seventh cycle.
- Q 1,2 , Q 2,2 and Q 3,2 stored in the memory 30 are updated to Q′ 1,2 , Q′ 2,2 and Q′ 3,2 .
- the final result of the addition process means the sum of ⁇ (q i,j ) values corresponding to positions in the i th row of the check matrix H, at which the bit is set to “1”.
- ⁇ (q i,j4 ) is subtracted from the final result of the addition-process.
- the multiplexer 112 b of the subtraction circuit 112 selects the one of the registers 111 a that holds the final result of the addition process.
- the multiplexer 112 a of the subtraction circuit 112 selects ⁇ (q i,j4 ).
- the subtracter 112 c of the subtraction circuit 112 subtracts ⁇ (q i,j4 ) selected by the multiplexer 112 a from the final result of the addition process selected by the multiplexer 112 b.
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M excludes MSB(q i,j4 ) from the final result of the exclusive-OR computation.
- the multiplexer 136 in the sign computation circuit 130 selects the one of the registers 131 that holds the final result of the exclusive-OR computation.
- the multiplexer 135 in the sign computation circuit 130 selects MSB(q i,j4 ).
- the exclusive-OR circuit 137 of the sign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by the multiplexer 136 and MSB(q i,j4 ) selected by the multiplexer 135 .
- bit information r i,j is acquired.
- the bit information r i,j is stored in the memory 30 at a position determined from the i th row and j th column corresponding to the bit information r i,j .
- R 1,4 , R 2,4 and R 3,4 are generated and stored in the memory 30 .
- R 1,4 , R 2,4 and R 3,4 represent a set of r i,j values corresponding to the bit positions in the permuted matrices I 1,4 , I 2,4 and I 3,4, which correspond to the (3m+1) th to 4m th columns (of the fourth column block) included in the check matrix H, and at which the bit is set to “1”.
- the parity-processing units 20 - 1 to 20 - m perform parity update computation, utilizing each element r i,j in R 1,3 , R 2,3 and R 3,3 stored in the memory 30 in the eighth cycle.
- Q 1,3 , Q 2,3 and Q 3,3 stored in the memory 30 are updated to Q′ 1,3 , Q′ 2,3 and Q′ 3,3 .
- the final result of the addition process means the sum of ⁇ (q i,j ) values corresponding to positions in the i th row of the check matrix H, at which the bit is set to “1”.
- ⁇ (q i,j5 ) is subtracted from the final result of the addition process.
- the multiplexer 112 b of the subtraction circuit 112 selects the one of the registers 111 a that holds the final result of the addition process.
- the multiplexer 112 a of the subtraction circuit 112 selects ⁇ (q i,j5 ).
- the subtracter 112 c of the subtraction circuit 112 subtracts ⁇ (q i,j5 ) selected by the multiplexer 112 a from the final result of the addition process selected by the multiplexer 112 b.
- the sign computation circuit 130 of each of the bit-processing units 10 - 1 to 10 -M excludes MSB(q i,j5 ) from the final result of the exclusive-OR computation.
- the multiplexer 136 in the sign computation circuit 130 selects the one of the registers 131 that holds the final result of the exclusive-OR computation.
- the multiplexer 135 in the sign computation circuit 130 selects MSB(q i,j5 ).
- the exclusive-OR circuit 137 of the sign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by the multiplexer 136 and MSB(q i,j5 ) selected by the multiplexer 135 .
- bit information r i,j is acquired.
- the bit information r i,j is stored in the memory 30 at a position determined from the i th row and j th column corresponding to the bit information r i,j .
- R 1,5 , R 2,5 and R 3,5 are generated and stored in the memory 30 .
- R 1,5 , R 2,5 and R 3,5 represent a set of r i,j values corresponding to the bit positions in the permuted matrices I 1,5 , I 2,5 and I 3,5, which correspond to the (4m+1) th to 5m th (N th ) columns (of the fifth column block) included in the check matrix H, and at which the bit is set to “1”.
- the parity-processing units 20 - 1 to 20 - m perform parity update computation, utilizing each element r i,j in R 1,4 , R 2,4 and R 3,4 stored in the memory 30 in the ninth cycle.
- Q 1,4 , Q 2,4 and Q 3,4 stored in the memory 30 are updated to Q′ 1,4 , Q′ 2,4 and Q′ 3,4 .
- bit-processing units 10 - 1 to 10 -M perform the same addition process as in the fourth cycle, using Q′ 1,4 , Q′ 2,4 and Q′ 3,4 .
- the parity-processing units 20 - 1 to 20 - m perform parity update computation using each element r i,j in R 1,5 , R 2,5 and R 3,5 stored in the memory 30 in the tenth cycle.
- Q 1,5 , Q 2,5 and Q 3,5 stored in the memory 30 are updated to Q′ 1,5 , Q′ 2,5 and Q′ 3,5 , which is the completion of the first iteration operation including bit update computation and parity update computation.
- bit update computation included in the second iteration operation is already started in the eighth cycle (i.e., in the cycle following the cycle in which parity update computation of the preceding (first) iteration operation is started).
- each parity-processing unit 20 - k is configured to simultaneously perform parity update computation concerning all positions (i th row position) in the j th column of the check matrix H, using corresponding r i,j .
- a parity-processing unit that has the same structure as each bit-processing unit 10 - i may be employed instead of each parity-processing unit 20 - k .
- parity information q i,j corresponding to the i th row position in the j th column of the check matrix H- can be sequentially updated by an addition process and subtraction process in two cycles. In this case, it is sufficient if only a single lookup table is provided at the output side of the parity-processing unit.
- the check matrix H is formed of (3 ⁇ 5) permuted matrices.
- the arrangement of the permuted matrices in the check matrix H is not limited to the above. It is sufficient if the check matrix H is formed of (r ⁇ s) permuted matrices, assuming that r ⁇ s.
Landscapes
- Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Error Detection And Correction (AREA)
- Detection And Correction Of Errors (AREA)
- Detection And Prevention Of Errors In Transmission (AREA)
Abstract
Description
- This application is based upon and claims the benefit of priority from prior Japanese Patent Application No. 2004-193767, filed Jun. 30, 2004, the entire contents of which are incorporated herein by reference.
- 1. Field of the Invention
- The present invention relates to a decoder and decoding method for decoding low-density parity-check codes with parity check matrices.
- 2. Description of the Related Art
- Attention has recently been paid to low-density parity-check codes (LDPC codes) used as error detection and correction codes. Message-passing algorithms are known as a technique for decoding LDPC codes from, for example, Andrew J. Blanksby and Chris J. Howland, “A 690-mW 1-Gb/s 1024-b, Rate-½ Low-Density Parity-Check Code Decoder”, IEEE Journal of Solid-State Circuits, Vol. 37, No. 3, pp. 404-412, March 2002. In the message-passing algorithms, to decode LDPC codes, bit update computation, which is performed for each row of a check matrix (parity-check matrix), and parity update computation, which is performed for each column of the matrix, are iteratedly executed.
- When decoding LDPC codes using the message-passing algorithms, bit update computation and parity update computation depend upon each other. Therefore, in the prior art, parity update computation of one loop cannot be started unless bit update computation of one loop is finished, and bit update computation of the next loop cannot be started unless parity update computation of the one loop is finished.
- In accordance with an embodiment of the invention, there is provided a low-density parity-check code decoder. The low-density parity-check code decoder comprises (r×m) bit-processing units, m parity-processing units and a controller. The (r×m) bit-processing units are configured to perform bit update computation for sequentially updating bit information corresponding to column positions included in the respective rows of a parity-check matrix, the parity-check matrix being formed of (r×s) permuted matrices each having an array of (m×m) and being divided into s column blocks, a bit at each of the column positions being set to “1”. The m parity-processing units are configured to perform parity update computation for updating parity information corresponding to row positions in m columns of each column block of the parity-check matrix, a bit at each of the row positions being set to “1”. The controller is configured to cause, whenever the (r×m) bit-processing units have finished bit update computation for m column positions in the respective rows of the parity-check matrix, the m parity-processing units to perform parity update computation corresponding to m columns of one of the s column blocks to which the m column positions belong. Further, the controller is configured to cause, after the parity-processing units finish parity update computation for m columns of a first one of the s column blocks, the (r×m) bit-processing units to start next bit update computation.
- The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention, and together with the general description given above and the detailed description of the embodiments given below, serve to explain the principles of the invention.
-
FIG. 1 is a view illustrating the structure of a parity-check matrix H employed in an embodiment of the invention; -
FIG. 2 is a block diagram illustrating the configuration of an LDPC-code decoder according to the embodiment; -
FIG. 3 is a block diagram illustrating the configuration of the bit-processing unit 10-i (i=1, 2, . . . , M) appearing inFIG. 2 ; -
FIG. 4 is a block diagram illustrating the configuration of the parity-processing unit 20-k (k=1, 2, . . . , m) appearing inFIG. 2 ; -
FIG. 5 is a view illustrating information input and output in each cycle during bit update computation by the bit-processing units 10-1 to 10-M and parity update computation by the parity-processing units 20-1 to 20-m; -
FIG. 6 is a view illustrating changes in the states of the areas of amemory 30 corresponding to respective (3×5 (r×s)) permuted matrices that provide a check matrix H; and -
FIG. 7 is a view useful in explaining a schedule for decoding an LDPC code employed in the embodiment. - An embodiment of the invention will be described with reference to the accompanying drawings. Firstly, a description will be given of a parity-check matrix H of low-density parity-check codes (LDPC codes) employed in the embodiment. As shown in
FIG. 1 , the parity-check matrix H is an array of (r×s) permuted matrices Ig,h (g=1, 2, . . . , r; h=1, 2, . . . , s). Each permuted matrix Ig,h is a sub-matrix of the check matrix H. Each permuted matrix Ig,h is acquired by iterating, a number of times, exchange of two different rows or columns included in unit matrices I each having an array of (m×m). In this case, the number M of rows of the check matrix H is rm, and the number N of columns of the matrix H is sm. In the example ofFIG. 1 , r=3 and s=5. Further, the check matrix H is divided into s column blocks (i.e., the 1st to the sth column blocks). Permuted matrices I1,h, I2,h and I3,h belong to the hth column block. - The check matrix H contains M (=rm) parity nodes, and N (=sm) bit nodes. This check matrix H is constructed, assuming that the LDPC code is N (=sm) bits. The jth bit (j=1, 2, . . . , N) of the LDPC code is represented by xj. As already known, each row of the check matrix H can be regarded as a parity-check equation. The parity-check equation corresponding to the ith row of the check matrix H is called an ith parity-check equation. The N-bit LDPC code must satisfy all parity-check equations of the check matrix H, i.e., the first to the Mth parity-check equations.
- The oblique lines in each permutated matrix Ig,h indicate the locus of the positions of “1” bits. As is evident from
FIG. 1 , only one “1” bit exists in each row and each column of {3(r)×5(s)} permutated matrices that provide the check matrix H. Accordingly, the numbers of “1” bits contained in each row and each column of the check matrix H are s and r, respectively. r and s are set to satisfy r<s. Accordingly, the number of “1” bits in each row is greater than that in each column. - To decode each LDPC code using the message-passing algorisms, it is necessary, in general, to iterate bit update computation and parity update computation.
- Generally, bit update computation can be expressed by the following equation:
- In the above equation, ri,j is acquired by passing corresponding parity information (reliability information) qi,j to each bit of the ith row of the check matrix H (except for the jth bit in the ith row). ri,j is bit information (reliability information) indicating reliability concerning code bit xj corresponding to the jth bit in the ith row. Assume here that ri,j does not indicate the probability itself that code bit xj is 0 or 1, but the logarithmic value of the probability. The reason why such a logarithmic value is used is that it enables multiplication/division to be replaced with addition/subtraction.
- M(i) represents a set of column positions of “1” bits in the ith row of the check matrix H. M(i)\j represents M(i) except for j. j′ε M(i)\j represents an arbitrary column position included in M(i) except for j. qi,j′ is parity information (reliability information) indicating reliability concerning code bit xj, corresponding to the jth bit in the ith row of the check matrix H. Assume that qi,j′ does not indicate the probability itself that code bit xj′ is 0 or 1, but the logarithmic value of the probability.
- sgn (qi,j′) represents the sign of qi,j′. The sign of qi,j′ is represented by the most significant bit MSB of qi,j′. If qi,j′ is negative, sgn (qi,j′)=−1, and MSB (qi,j′)=1. In contrast, if qi,j′ is not negative, i.e., if it is 0 or positive, sgn (qi,j′)=+1, and MSB (qi,j′)=0. ψ( ) represents a function (hereinafter referred to as “ψ function”) for probability computation.
- Further, parity update computation is expressed by the following equation:
- In the above equation (2), qi,j is acquired by passing, from a certain bit node to a corresponding parity node, bit information (reliability information) ri′,j corresponding to each bit of the jth column of the check matrix H (except for the ith bit in the jth column). qi,j is parity information (reliability information) indicating reliability concerning code bit xj corresponding to the jth bit in the ith row of the check matrix H. qi,j is used as the above-mentioned qi,j′ for bit update computation expressed by the equation (1).
- pj represents the logarithmic value of the probability (initial probability) that the jth bit xj is 1 or 0, which is estimated from an LDPC code when it is output through a communication channel. The initial value of qi,j corresponding to each “1” bit in the check matrix H is identical to pj.
- N(j) represents a set of row positions of “1” bits in the jth column-of the check matrix H. N(j)\i represents N(j) except for i. i′ε N(j)\i represents an arbitrary row position included in N(j) except for i. ri′,j is bit information (reliability information) indicating reliability concerning the jth bit xj in the ith row of the check matrix H. ri,j updated by computation using the equation (1) is used as ri′,j.
-
FIG. 2 is a block diagram illustrating the configuration of an LDPC-code decoder according to the embodiment. As shown inFIG. 2 , the LDPC-code decoder comprises M (=rm=3m) bit-processing units 10-1 to 10-M, m parity-processing units 20-1 to 20-m, amemory 30 and acontroller 40. The bit-processing units 10-1 to 10-M are provided for the first to Mth rows of the check matrix H, respectively. Each bit-processing unit 10-i (i=1, 2, . . . , M) executes bit update computation, expressed by the equation (1), on the ith row of the check matrix H. - The parity-processing units 20-1 to 20-m are common units used for all permuted matrices Ig,h, and provided for the first to mth columns of each permuted matrix Ig,h. Concerning h (=1, 2, . . . , s; s=5 in the embodiment), each parity-processing unit 20-k (k=1, 2, . . . m) is used to perform parity update computation on the ((h-1)m+k)th column in the check matrix H, to which the kth column of each permuted matrix Ig,h belongs. Namely, in the embodiment wherein s=5, each parity-processing unit 20-k is used to perform parity update computation on the kth column, (m+k)th column, (2m+k)th column, (3m +k)th column and (4m +k)th column of the check matrix H. The parity-processing unit 20-1, for example, is used to perform parity update computation on the first column, (m+1)th column, (2m+1)th column, (3m+1)th column and (4m+1)th column of the check matrix H. Similarly, the parity-processing unit 20-m, for example, is used to perform parity update computation on the mth column, 2mth column, 3mth column, 4mth column and 5mth column (the Nth column) of the check matrix H.
- The
memory 30 is used to store bit information and parity information. The bit information is calculated by the bit-processing units 10-1 to 10-M. The parity information is calculated by the parity-processing units 20-1 to 20-m. Assume here that thememory 30 is formed of (3×5) (=r×s) memory units corresponding to (3×5) (=r×s) permuted matrices included in the check matrix H, i.e., I1,1, I2,1, I3,1, I1,2, I2,2, I3,2, I1,3, I2,3, I3,3, I1,4, I2,4, I3,4, I1,5, I2,5 and I3,5. The (3×5) memory units can be simultaneously accessed. The state of the storage area of thememory 30 corresponding to each of permuted matrices Ig,h is represented by Qg,h or Rg,h. Qg,h indicates that parity information is stored in a storage area of thememory 30 corresponding to Qg,h. Rg,h indicates that bit information is stored in a storage area of thememory 30 corresponding to Rg,h. - The
controller 40 controls decoding processing of an LDPC code input to the LDPC-code decoder. Specifically, thecontroller 40 functions as a sequencer for controlling the operations of the bit-processing units 10-1 to 10-M and parity-processing unit 20-1 to 20-m. -
FIG. 3 is a block diagram illustrating the bit-processing unit 10-i (i=1, 2, . . . , M) appearing inFIG. 2 . Bit update computation expressed by the equation (1) mainly comprises a product part of H operation, a ψ function part and a summation part of Σ operation. The ψ function part, i.e., the ψ( ) operation, can be realized using a lookup table. In light of this, the bit-processing unit 10-i is formed of asummation computation circuit 110, lookup table (LUT) 120 and signcomputation circuit 130. Thesummation computation circuit 110 corresponds to the summation part. The lookup table 120 and signcomputation circuit 130 correspond to the ψ function part and product part, respectively. - The summation part can be divided into an adder section and subtracter section. The adder section calculates the sum of ψ(qi,j) values corresponding to positions j in the ith row of the check matrix H, at which the bit is set to “1”. The subtracter section subtracts, from the sum calculated by the adder section, ψ(qi,j) corresponding to position j in the ith row at which the bit is set to “1”. In light of this, in the embodiment, the
summation computation circuit 110 is formed of anaddition circuit 111 andsubtraction circuit 112. - The
addition circuit 111 comprises a pair ofregisters 111 a, multiplexers (MUX) 111 b and 111 b, andadder 111 d. Theregisters 111 a are used to hold the intermediate result (or final result) of addition processing for calculating the sum of ψ(qi,j) values. The final result of addition processing is the last intermediate result, and indicates the sum of ψ(qi,j) values. When one of theregisters 111 a holds the sum of ψ(qi,j) values, theother register 111 a is used to hold a new intermediate result of addition processing used for calculating a new sum of ψ(qi,j) values. Thus, the functions of theregisters 111 a as respective registers for holding the sum and the intermediate result of ψ(qi,j) values are switched each time the sum of ψ(qi,j) values is calculated. - The
multiplexer 111 b sequentially selects ψ(qi,j) corresponding to each position (column position) j in the ith row of the check matrix H, at which the bit is set to “1”. Position j in the ith row of the check matrix H, at which the bit is set to “1”, can be determined uniquely in units of rows. Accordingly, ψ(qi,j) corresponding to position j in the ith row at which the bit is set to “1” can be determined uniquely. The multiplexer 111 c selects the intermediate result (the result of the preceding addition process) held by one of theregisters 111 a. Theadder 111 d adds (qi,j) selected by themultiplexer 111 b to the intermediate result selected by the multiplexer 111 c. The addition result of theadder 111 d indicates a new intermediate result of addition processing for calculating the sum of ψ(qi,j) values. The intermediate result held by theregister 111 a and used for addition of theadder 111 d is updated to the addition result of theadder 111 d, i.e., the new intermediate result of addition processing for calculating the sum of ψ(qi,j) values. - As described above, the
multiplexer 111 b sequentially selects ψ(qi,j) corresponding to each position j in the ith row of the check matrix H, at which the bit is set to “1”. Further, each time themultiplexer 111 b selects ψ(qi,j), theaddition circuit 111 adds the selected ψ(qi,j) to the intermediate result held by one of theregisters 111 a (i.e., the intermediate result selected by the multiplexer 111 c). Each time this addition is performed, the intermediate result of theregister 111 a used for the addition is updated to a new intermediate result as the addition result. Assume that later on, ψ(qi,j) corresponding to the last position j in the ith row is added to the intermediate result held by theregister 111 a at this time. As is apparent, the result of this addition indicates the sum of ψ(qi,j) values, i.e., the sum of ψ(qi,j) values corresponding to all positions j included in M(i). - Assume further that processing of one row of each permuted matrices Ig,h in the
addition circuit 111, i.e., one addition process, can be performed by one cycle. In this case, the operation of theaddition circuit 111 for calculating the sum of ψ(qi,j) values can be executed in s (s=5) cycles. Assuming that one cycle is time Tcyc, the sum of ψ(qi,j) values can be executed in s Tcyc (=5 Tcyc). - The
subtraction circuit 112 comprises multiplexers (MUX) 112 a and 112 b, and a subtracter 112 c. Themultiplexer 112 a sequentially selects ψ(qi,j) corresponding to each position j in the ith row of the check matrix H, at which the bit is set to “1”, in cycles following the cycle in which the sum of ψ(qi,j) values corresponding to positions j in the ith row of the check matrix H, at which the bit is set to “1”, is held by one of theregisters 111 a. On the other hand, themultiplexer 112 b selects the output of theregister 111 a, i.e., the above-mentioned sum. The subtracter 112 c subtracts ψ(qi,j) selected by themultiplexer 112 a from the sum selected by themultiplexer 112 b. As a result, the sum of ψ(qi,j) values corresponding to all positions j′ that satisfy M(i)\j can be acquired. The subtraction, by the subtracter 112 c, of ψ(qi,j) from the sum held by the oneregister 111 a is performed concerning ψ(qi,j) values corresponding to all positions j in the ith row of the check matrix H, at which the bit is set to “1”. - Thus, the
subtraction circuit 112 sequentially calculates the sum of ψ(qi,j′) values corresponding to all positions j′ that satisfy M(i)\j. The calculation of the sum of ψ(qi,j′) values corresponding to all positions j′ that satisfy M(i)\j, performed concerning each position j in the ith row of the check matrix H, at which the bit is set to “1”, can be executed in s (=5) cycles. During the subtraction process by thesubtraction circuit 112, theaddition circuit 111 can perform the next addition process. At this time, the multiplexer 111 c selects the other of theregisters 111 a (i.e., the register that is not selected by themultiplexer 112 b during the subtraction process by the subtraction circuit 112). The content of the other register is updated to the addition result of theadder 111 d included in theaddition circuit 111. However, the next addition process by theaddition circuit 111 is started after the parity-processing unit 20-k acquires new ψ(qi,j) corresponding to each new position i in the first to mth rows of the check matrix H, at which the bit is set to “1”. - The lookup table 120 is used to convert the subtraction result of the subtracter 112 (i.e., the computation result of the sum computation circuit 110) into an operation value of the ψ function, i.e., ψ′ (subtraction result). The lookup table 120 is referred to, using the subtraction result of the
subtracter 112. Assume that the subtraction result is x. The entry of the lookup table 120 designated by x prestores ψ(x) unique to x. Accordingly, ψ(x) can be acquired from the lookup table 120 simply by referring to the table 120 using the subtraction result x of thesubtracter 112. In the embodiment, the entry of the lookup table 120 designated by x stores ψ(x)=-log(tanh(x/2)). - The
sign computation circuit 130 comprises a pair ofregisters 131, multiplexers (MUX) 132 and 133, exclusive-OR circuit 134, multiplexers (MUX) 135 and 136, exclusive-OR circuit 137 and multiplexer (MUX) 138. Theregisters 131 are each a register of, for example, 1 bit used to hold the intermediate result (or final result) of sign computation. When one of theregisters 131 holds the final result of sign computation, the other register is used to hold the intermediate result of new sign computation. - The
multiplexer 132 sequentially selects an MSB of qi,j, i.e., a sign, corresponding to each position j in the ith row of the check matrix H, at which the bit is set to “1”. Each MSB of qi,j will be often expressed by MSB(qi,j). Themultiplexer 133 selects the intermediate result (sign) of sign computation so far held by one of theregisters 111 a. The exclusive-OR circuit 134 acquires the exclusive-OR of the sign selected by themultiplexer 132 and the intermediate result (sign) of sign computation so far. The exclusive-OR result of the exclusive-OR circuit 134 indicates a new intermediate result of sign computation. The intermediate result in theregister 131 used for the operation of the exclusive-OR circuit 134 is updated to the exclusive-OR result of the exclusive-OR circuit 134, i.e., the new intermediate result of sign computation. - As described above, the
multiplexer 132 sequentially selects MSB(qi,j) corresponding to each position j in the ith row of the check matrix H, at which the bit is set to “1”. Each time themultiplexer 132 selects MSB(qi,j), thesign computation circuit 130 executes the operation of acquiring the exclusive-OR of the selected MSB(qi,j) and the intermediate result (sign) of sign computation held by theregister 131 at this time. Further, each time exclusive-OR is computed, the intermediate result held by theregister 131 and used for the exclusive-OR computation is updated to a new intermediate result as the result of the computation. Assume that later on, the exclusive-OR of MSB(qi,j) corresponding to the last position j in the ith row and the intermediate result held by theregister 131 at this time is computed. As is apparent, the result (final result) of this exclusive-OR computation indicates the product of the signs of qi,j corresponding to all positions j in the ith row at which the bit is set to “1”, i.e., the product of the signs of qi,j corresponding to all positions j included in M(i). The above-described operation of thesign computation circuit 130 for computing the exclusive-OR concerning MSB(qi,j), namely, the sign computation, is performed in synchronism with the addition process of theaddition circuit 111, and can be executed in s (=5) cycles. - The
multiplexer 135 sequentially selects MSB(qi,j) corresponding to each position j in the ith row of the check matrix H, at which the bit is set to “1”. The selection operation of themultiplexer 135 is executed in cycles following the cycle in which the exclusive-OR concerning MSB(qi,j) values corresponding to all positions j included in M(i) (i.e., the final result of the exclusive-OR computation indicating the product) is held by one of theregisters 131. On the other hand, themultiplexer 136 selects the output of theregister 131, i.e., the above-mentioned exclusive-OR (the final result of the exclusive-OR computation). The exclusive-OR circuit 137 acquires the exclusive-OR of the exclusive-OR selected by themultiplexer 136 and MSB(qi,j) selected by themultiplexer 135. As a result, the product of the signs of qi,j′ values corresponding to all positions j′ that satisfy M(i)\j can be acquired. The operation of the exclusive-OR circuit 137 for acquiring the exclusive-OR of the product (held by the one register 131) and MSB(qi,j) is performed concerning MSB(qi,j) values corresponding to all positions j in the ith row of the check matrix H, at which the bit is set to “1”. Accordingly, concerning each position j in the ith row of the check matrix H, at which the bit is set to “1”, thesign computation circuit 130 sequentially computes the product of MSB(qi,j) values corresponding to all positions j′ that satisfy M(i)\j. In accordance with the output of the exclusive-OR circuit 137, themultiplexer 138 selects, as ri,j, the output of the lookup table 120 or a value acquired by inverting the sign of this output. -
FIG. 4 is a block diagram illustrating the configuration of the parity-processing unit 20-k (k=1, 2, . . . , m). The parity-processing unit 20-k is constructed in light of the feature that the number (3 in the embodiment) of “1” bits in each column of the check matrix H is smaller than the number (5 in the embodiment) of “1” bits in each row of the check matrix H. Namely, the parity-processing unit 20-k is configured to simultaneously perform parity update computation on all columns j, using ri,j corresponding to each position (row position) i in the jth column of the check matrix H, at which the bit is set to “1”. To this end, the parity-processing unit 20-k comprises threeadders subtracters FIG. 4 . - The
adder 201 adds ri1,j to ri2,j, and theadder 202 adds ri3,j to pj. ri1,j, ri2,j and ri3,j represent bit information items corresponding to bit positions i1, i2 and i3 in the jth column (column j) of the check matrix H, at which the bit is set to “1”. Assume here that the jth column of the check matrix H is positioned in three permuted matrices I1,h, I2,h and I3,h (of the hth column block). In this case, each of the permuted matrices I1,h, I2,h and I3,h contains one position in the jth column, at which the bit is set to “1”. Namely, ri1,j, ri2,j and ri3,j represent bit information items corresponding to bit positions in the permuted matrices I1,h, I2,h and I3,h and in the jth column of the check matrix H, at which the bit is set to “1”. The bit positions i1, i2 and i3 in the jth column of the check matrix H, at which the bit is set to “1”, can be uniquely determined in units of rows. Accordingly, ri1,j, ri2,j and ri3,j corresponding to the bit positions i1, i2 and i3 in the jth column, at which the bit is set to “1”, can be uniquely determined. - The
adder 203 adds up the addition results of theadders adder 203 sums up pj and the total of ri1,j, ri2,j and ri3,j. Thus, theadders 201 to 203 provide an addition circuit for summing up pj and the total of ri1,j, ri2,j and ri3,j. The addition result of theadder 203 indicates the probability qj of bit xj in the LDPC code being 1 or 0, which is estimated at the present stage. - The
subtracter 204 subtracts ri1,j from the addition result of theadder 203. The subtraction result of thesubtracter 204 indicates updated parity information qi1,j. Thesubtracter 205 subtracts ri2,j from the addition result of theadder 203. The subtraction result of thesubtracter 205 indicates updated parity information qi2,j. Thesubtracter 206 subtracts ri3,j from the addition result of theadder 203. The subtraction result of thesubtracter 206 indicates updated parity information qi3,j. In other words, in each parity-processing unit 20-k, three qi1,j, qi2,j and qi3,j corresponding to “1” bits in the jth column (j=k, m+k, 2m+k, 3m+k, 4m+k) of the check matrix H can be simultaneously acquired. As a result, MSB(qi1,j), MSB(qi2,j) and MSB(qi3,j) can be acquired. Further, ψ(qi1,j), ψ(qi2,j) and ψ(qi3,j) can also be simultaneously acquired by simultaneously referring to the lookup tables 207, 208 and 209 using qi1,j, qi2,j and qi3,j as the subtraction results of thesubtracters - As described above, one feature of the embodiment lies in that the lookup tables 207, 208 and 209 are provided at the output side of each parity-processing unit 20-k. Instead of providing the lookup tables 207, 208 and 209 at the output side of each parity-processing unit 20-k, respective lookup tables can be provided at the output side of the
multiplexers - Referring then to
FIGS. 5 and 6 , a description will be given of the flow of computation performed in the bit-processing units 10-1 to 10-M and parity-processing units 20-1 to 20-m. The bit-processing units 10-1 to 10-M and parity-processing units 20-1 to 20-m are controlled by thecontroller 40 incorporated in the LDPC-code decoder shown inFIG. 2 .FIG. 5 shows input/output information in each cycle during bit update computation by the bit-processing units 10-1 to 10-M and parity update computation by the parity-processing units 20-1 to 20-m. The bit update computation and parity update computation may hereinafter be referred to as “horizontal calculation HCALC” and “vertical calculation VCALC”, respectively.FIG. 6 shows changes in the states of regions that are included in thememory 30 and correspond to the (r×s=15) permuted matrices of the check matrix H. InFIGS. 5 and 6 , a set of ψ(qi,j) values and a set of ri,j values are represented by Qg,h and Rg,h, respectively. Qg,h includes MSB(qi,j). - <Initialization>
- Firstly, the
controller 40 executes an initialization process. In the initialization process, thecontroller 40 stores ψ(pj) and sgn(pj) into each memory unit of thememory 30. pj represents a logarithmic value of probability acquired from the output of the communication channel. Concerning all j values, ψ(pj) is set as the initial value of each of ψ(qi1,j), ψ(qi2,j) and ψ(qi3,j). qi1,j, qi2,j and qi3,j represent parity information items qi,j (i=i1, i2, i3) corresponding to bit positions i1, i2 and i3 in the jth column of the check matrix H, at which the bit is set to “1”. Further, in the initialization process, sgn(pj) is set as the initial value of each of MSB(qi1,j), MSB(qi2,j) and MSB(qi3,j). Thecontroller 40 also resets theregisters 31 and 111 a incorporated in each of the M bit-processing units 10-1 to 10-M. - <First Cycle>
- In the first cycle, the M bit-processing units 10-1 to 10-M start addition processing. Specifically, MSB(qi,j1) and ψ(qi,j1) are stored into the
registers registers memory 30. The first bit position ji belongs to the first column block. - <Second Cycle>
- In the second cycle, the
multiplexer 111 b in theaddition circuit 111 of each of the bit-processing units 10-1 to 10-M selects ψ(qi,j2). Theaddition circuit 111 performs addition processing, using ψ(qi,j2). Specifically, theadder 111 d of theaddition circuit 111 adds up the value stored in theregister 111 a in the first cycle, and ψ(qi,j2) selected by themultiplexer 111 b. The addition result of theadder 111 d, i.e., the sum of the value (ψ(qi,j1)) stored in theregister 111 a and ψ(qi,j2), is stored into theregister 111 a. On the other hand, thesign computation circuit 130 of each of the bit-processing units 10-1 to 10-M selects MSB(qi,j2), using themultiplexer 132. In the second cycle, using MSB(qi,j2), the exclusive-OR circuit 134 of thesign computation circuit 130 executes exclusive-OR computation. Namely, the exclusive-OR circuit 134 calculates the exclusive-OR of the value (MSB(qi,j1)) stored in theregister 131 in the first cycle, and MSB(qi,j2) selected by themultiplexer 132. The computation result of the exclusive-OR circuit 134, i.e., the exclusive-OR of the value stored in theregister 131 and MSB(qi,j2), is stored in theregister 131. MSB(qi,j2) and ψ(qi,j2) correspond to the second bit position j2 (j=j2) included in the bit positions j1 to j5 in each (ith) row of the check matrix H at which the bit is set to “1”, and correspond to Q1,2, Q2,2 and Q3,2 stored in thememory 30. The second bit position j2 belongs to the second column block. - <Third Cycle>
- In the third cycle, the
multiplexer 111 b in theaddition circuit 111 of each of the bit-processing units 10-1 to 10-M selects ψ(qi,j3). Theaddition circuit 111 performs addition processing, using ψ(qi,j3). Specifically, theadder 111 d of theaddition circuit 111 adds up the value stored in theregister 111 a in the second cycle, and ψ(qi,j3) selected by themultiplexer 111 b. The addition result of theadder 111 d, i.e., the sum of the value stored in theregister 111 a and ψ(qi,j3), is stored into theregister 111 a. On the other hand, thesign computation circuit 130 of each of the bit-processing units 10-1 to 10-M selects MSB(qi,j3), using themultiplexer 132. In the third cycle, using MSB(qi,j3), the exclusive-OR circuit 134 of thesign computation circuit 130 executes exclusive-OR computation. Namely, the exclusive-OR circuit 134 calculates the exclusive-OR of the value stored in theregister 131 in the second cycle, and MSB(qi,j3) selected by themultiplexer 132. The computation result of the exclusive-OR circuit 134, i.e., the exclusive-OR of the value stored in theregister 131 and MSB(qi,j3), is stored in theregister 131. MSB(qi,j3) and ψ(qi,j3) correspond to the third bit position j3 (j=j3) included in the bit positions j1 to j5 in each (ith) row of the check matrix H at which the bit is set to “1”, and correspond to Q1,3, Q2,3 and Q3,3 stored in thememory 30. The third bit position j3 belongs to the third column block. - <Fourth Cycle>
- In the fourth cycle, the
multiplexer 111 b in theaddition circuit 111 of each of the bit-processing units 10-1 to 10-M selects ψ(qi,j4). Theaddition circuit 111 performs addition processing, using ψ(qi,j4). Specifically, theadder 111 d of theaddition circuit 111 adds up the value stored in theregister 111 a in the third cycle, and ψ(qi,j 4) selected by themultiplexer 111 b. The addition result of theadder 111 d, i.e., the sum of the value stored in theregister 111 a and ψ(qi,j4), is stored into theregister 111 a. On the other hand, thesign computation circuit 130 of each of the bit-processing units 10-1 to 10-M selects MSB(qi,j4), using themultiplexer 132. In the fourth cycle, using MSB(qi,j4), the exclusive-OR circuit 134 of thesign computation circuit 130 executes exclusive-OR computation. Namely, the exclusive-OR circuit 134 calculates the exclusive-OR of the value stored in theregister 131 in the third cycle, and MSB(qi,j4) selected by themultiplexer 132. The computation result of the exclusive-OR circuit 134, i.e., the exclusive-OR of the value stored in theregister 131 and MSB(qi,j4), is stored in theregister 131. MSB(qi,j4) and (qi,j4) correspond to the fourth bit position j4 (j=j4) included in the bit positions j1 to j5 in each (ith) row of the check matrix H at which the bit is set to “1”, and correspond to Q1,4, Q2,4 and Q3,4 stored in thememory 30. The fourth bit position j4 belongs to the fourth column block. - <Fifth Cycle>
- In the fifth cycle, the
multiplexer 111 b in theaddition circuit 111 of each of the bit-processing units 10-1 to 10-M selects *(qi,j5). Theaddition circuit 111 performs addition processing, using ψ(qi,j5). Specifically, theadder 111 d of theaddition circuit 111 adds up the value stored in theregister 111 a in the fourth cycle, and ψ(qi,j5) selected by themultiplexer 111 b. The addition result of theadder 111 d, i.e., the sum of the value stored in theregister 111 a and ψ(qi,j5), is stored into theregister 111 a. On the other hand, thesign computation circuit 130 of each of the bit-processing units 10-1 to 10-M selects MSB(qi,j5), using themultiplexer 132. In the fifth cycle, using MSB(qi,j5), the exclusive-OR circuit 134 of thesign computation circuit 130 executes exclusive-OR computation. Namely, the exclusive-OR circuit 134 calculates the exclusive-OR of the value stored in theregister 131 in the fourth cycle, and MSB(qi,j5) selected by themultiplexer 132. The computation result of the exclusive-OR circuit 134, i.e., the exclusive-OR of the value stored in theregister 131 and MSB(qi,j5), is stored in theregister 131. MSB(qi,j5) and ψ(qi,j5) correspond to the fifth bit position j5 (j=j5) included in the bit positions j1 to j5 in each (ith) row of the check matrix H at which the bit is set to “1”, and correspond to Q1,5, Q2,5 and Q3,5 stored in thememory 30. The fifth bit position j5 belongs to the fifth column block. - <Sixth Cycle>
- In the sixth cycle, the
subtraction circuit 112 of each of the bit-processing units 10-1 to 10-M starts a subtraction process. The subtraction process utilizes the final result of the addition process held in one of theregisters 111 a in each bit-processing unit 10-i (i=1, 2, . . . , M), and Q1,1, Q2,1 and Q3,1 stored in thememory 30. The final result means the sum of ψ(qi,j) values corresponding to positions in the ith row of the check matrix H, at which the bit is set to “1”. In the subtraction process, ψ(qi,j1) is subtracted from the final result of the addition process. To this end, themultiplexer 112 b of thesubtraction circuit 112 selects the one of theregisters 111 a that holds the final result of the addition process. Themultiplexer 112 a of thesubtraction circuit 112 selects ψ(qi,j1). After that, the subtracter 112c of thesubtraction circuit 112 subtracts ψ(qi,j1) selected by themultiplexer 112 a from the final result of the addition process selected by themultiplexer 112 b. - On the other hand, the
sign computation circuit 130 of each of the bit-processing units 10-1 to 10-M excludes MSB(qi,j1) from the final result of the exclusive-OR computation. To this end, themultiplexer 136 in thesign computation circuit 130 selects the one of theregisters 131 that holds the final result of the exclusive-OR computation. The final result of the exclusive-OR computation means the exclusive-ORs of MSB(qi,j) values corresponding to positions j in the ith row of the check matrix H, at which the bit is set to “1”. Further, themultiplexer 135 in thesign computation circuit 130 selects MSB(qi,j1). After that, the exclusive-OR circuit 137 of thesign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by themultiplexer 136 and MSB(qi,j1) selected by themultiplexer 135. In accordance with the computation result of the exclusive-OR circuit 137, the sign of information output from themultiplexer 138 is determined, whereby bit information ri,j is acquired. The bit information ri,j is stored in thememory 30 at a position determined from the ith row and jth column corresponding to the bit information ri,j. In the sixth cycle, R1,1, R2,1 and R3,1 are generated and stored in thememory 30. R1,1, R2,1 and R3,1 represent a set of ri,j values corresponding to the bit positions in the permuted matrices I1,1, I2,1 and I3,1, which correspond to the 1st to mth columns (of the first column block) included in the check matrix H, and at which the bit is set to “1”. - <Seventh Cycle>
- In the seventh cycle, the
subtraction circuit 112 of each of the bit-processing units 10-1 to 10-M starts a subtraction process. The subtraction process utilizes the final result of the addition process held in one of theregisters 111 a in each bit-processing unit 10-i (i=1, 2, . . . , M), and Q1,2, Q2,2 and Q3,2 stored in thememory 30. The final result means the sum of ψ(qi,j) values corresponding to positions in the ith row of the check matrix H, at which the bit is set to “1”. In the subtraction process, ψ(qi,j2) is subtracted from the final result of the addition process. To this end, themultiplexer 112 b of thesubtraction circuit 112 selects the one of theregisters 111 a that holds the final result of the addition process. Themultiplexer 112 a of thesubtraction circuit 112 selects ψ(qi,j2). After that, the subtracter 112 c of thesubtraction circuit 112 subtracts ψ(qi,j2) selected by themultiplexer 112 a from the final result of the addition process selected by themultiplexer 112 b. - On the other hand, the
sign computation circuit 130 of each of the bit-processing units 10-1 to 10-M excludes MSB(qi,j2) from the final result of the exclusive-OR computation. To this end, themultiplexer 136 in thesign computation circuit 130 selects the one of theregisters 131 that holds the final result of the exclusive-OR computation. Further, themultiplexer 135 in thesign computation circuit 130 selects MSB(qi,j2). After that, the exclusive-OR circuit 137 of thesign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by themultiplexer 136 and MSB(qi,j2) selected by themultiplexer 135. In accordance with the computation result of the exclusive-OR circuit 137, the sign of information output from themultiplexer 138 is determined, whereby bit information ri,j is acquired. The bit information ri,j is stored in thememory 30 at a position determined from the ith row and jth column corresponding to the bit information ri,j. In the seventh cycle, R1,2, R2,2 and R3,2 are generated and stored in thememory 30. R1,2, R2,2 and R3,2 represent a set of ri,j values corresponding to the bit positions in the permuted matrices I1,2, I2,2 and I3,2, which correspond to the (m+1)th to 2mth columns (of the second column block) included in the check matrix H, and at which the bit is set to “1”. - In the seventh cycle, each element ri,j in R1,2 R2,2 and R3,2 stored in the
memory 30 can be utilized. Accordingly, in the seventh cycle, the parity-processing units 20-1 to 20-m start parity update computation under the control of thecontroller 40, utilizing each element ri,j in R1,2, R2,2 and R3,2. As a result, concerning each of the 1st to mth columns (of the first column block) included in the check matrix H, ψ(qi,j1), ψ(qi,j2) and ψ(qi,j3), MSB(qi,j1), MSB(qi,j2) and MSB(qi,j3), and qj are acquired. Using ψ(qi,j1), ψ(qi,j2) and ψ(qi,j3), MSB(qi,j1), MSB(qi,j2) and MSB(qi,j3), and qj acquired concerning each of the 1st to mth columns of the check matrix H, Q1,1, Q2,1 and Q3,1 stored in thememory 30 are updated to Q′1,1, Q′2,1 and Q′3,1. - Thus, in the embodiment, parity update computation can be started before bit update computation corresponding to all bits in each of the 1st to Mth rows of the check matrix H is finished. Namely, in the embodiment, parity update computation can be started when only bit update computation is finished which corresponds to each bit position in the permuted matrices I1,1, I2,1 and I3,1 (of the first column block) included in the check matrix H, at which the bit is set to “1”. In other words, in the embodiment, bit update computation and parity update computation can be processed using a pipeline.
- <Eighth Cycle>
- In the eighth cycle, each element qi,j in updated Q′1,2, Q′2,2 and Q′3,2 can be utilized. Accordingly, in the eighth cycle, the bit-processing units 10-1 to 10-M start next bit update computation under the control of the
controller 40, thereby performing the same addition process as in the first cycle. However, the result of the addition process is held by theregister 111 a other than theregister 111 a used to hold the result of the addition process of the first cycle (the first to fifth cycles). - Thus, in the embodiment, bit update computation can be started before parity update computation corresponding to all bits in each of the 1st to Nth columns of the check matrix H is finished. Namely, in the embodiment, next bit update computation can be started when only parity update computation is finished which corresponds to each bit position in the permuted matrices I1,1, I2,1 and I3,1 (of the first column block) included in the check matrix H (i.e., when parity update computation has been performed concerning the first m columns of the check matrix H). In other words, the combination of bit update computation and parity update computation can be iterated using a pipeline.
- Further, in the eighth cycle, a subtraction process by the bit-processing units 10-1 to 10-M is executed in parallel with the above addition process. In the subtraction process, the final result of the addition process held by the one of the
register 111 a in each of the bit-processing unit 10-1 to 10-M, and Q1,3, Q2,3 and Q3,3 stored in thememory 30 are used. The final result of the addition process means the sum of ψ(qi,j) values corresponding to positions in the ith row of the check matrix H, at which the bit is set to “1”. In the subtraction process, ψ(qi,j3) is subtracted from the final result of the addition process. To this end, themultiplexer 112 b of thesubtraction circuit 112 selects the one of theregisters 111 a that holds the final result of the addition process. Themultiplexer 112 a of thesubtraction circuit 112 selects ψ(qi,j3). After that, the subtracter 112 c of thesubtraction circuit 112 subtracts ψ(qi,j3) selected by themultiplexer 112 a from the final result of the addition process selected by themultiplexer 112 b. - On the other hand, the
sign computation circuit 130 of each of the bit-processing units 10-1 to 10-M excludes MSB(qi,j3) from the final result of the exclusive-OR computation. To this end, themultiplexer 136 in thesign computation circuit 130 selects the one of theregisters 131 that holds the final result of the exclusive-OR computation. Further, themultiplexer 135 in thesign computation circuit 130 selects MSB(qi,j3). After that, the exclusive-OR circuit 137 of thesign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by themultiplexer 136 and MSB(qi,j3) selected by themultiplexer 135. In accordance with the computation result of the exclusive-OR circuit 137, the sign of information output from themultiplexer 138 is determined, whereby bit information ri,j is acquired. The bit information ri,j is stored in thememory 30 at a position determined from the ith row and jth column corresponding to the bit information ri,j. In the eighth cycle, R1,3, R2,3 and R3,3 are generated and stored in thememory 30. R1,3, R2,3 and R3,3 represent a set of ri,j values corresponding to the bit positions in the permuted matrices I1,3, I2,3 and I3,3, which correspond to the (2m+1)th to 3mth columns (of the third column block) included in the check matrix H, and at which the bit is set to “1”. - Furthermore, in the eighth cycle, the parity-processing units 20-1 to 20-m perform parity update computation, utilizing each element ri,j in R1,2, R2,2 and R3,2 stored in the
memory 30 in the seventh cycle. As a result, Q1,2, Q2,2 and Q3,2 stored in thememory 30 are updated to Q′1,2, Q′2,2 and Q′3,2. - <Ninth Cycle>
- In the ninth cycle, the bit-processing units 10-1 to 10-M perform the same addition process as in the second cycle, using Q′1,2, Q′2,2 and Q′3,2. Further, in the ninth cycle, the bit-processing units 10-1 to 10-M perform a subtraction process in parallel with the addition process. In the subtraction process, the final result of the addition process held by the one of the
register 111 a in each bit-processing unit 10-i (i=1, 2, . . . , M), and Q1,4, Q2,4 and Q3,4 stored in thememory 30 are used. The final result of the addition process means the sum of ψ(qi,j) values corresponding to positions in the ith row of the check matrix H, at which the bit is set to “1”. In the subtraction process, ψ(qi,j4) is subtracted from the final result of the addition-process. To this end, themultiplexer 112 b of thesubtraction circuit 112 selects the one of theregisters 111 a that holds the final result of the addition process. Themultiplexer 112 a of thesubtraction circuit 112 selects ψ(qi,j4). After that, the subtracter 112 c of thesubtraction circuit 112 subtracts ψ(qi,j4) selected by themultiplexer 112 a from the final result of the addition process selected by themultiplexer 112 b. - On the other hand, the
sign computation circuit 130 of each of the bit-processing units 10-1 to 10-M excludes MSB(qi,j4) from the final result of the exclusive-OR computation. To this end, themultiplexer 136 in thesign computation circuit 130 selects the one of theregisters 131 that holds the final result of the exclusive-OR computation. Further, themultiplexer 135 in thesign computation circuit 130 selects MSB(qi,j4). After that, the exclusive-OR circuit 137 of thesign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by themultiplexer 136 and MSB(qi,j4) selected by themultiplexer 135. In accordance with the computation result of the exclusive-OR circuit 137, the sign of information output from themultiplexer 138 is determined, whereby bit information ri,j is acquired. The bit information ri,j is stored in thememory 30 at a position determined from the ith row and jth column corresponding to the bit information ri,j. In the ninth cycle, R1,4, R2,4 and R3,4 are generated and stored in thememory 30. R1,4, R2,4 and R3,4 represent a set of ri,j values corresponding to the bit positions in the permuted matrices I1,4, I2,4 and I3,4, which correspond to the (3m+1)th to 4mth columns (of the fourth column block) included in the check matrix H, and at which the bit is set to “1”. - Furthermore, in the ninth cycle, the parity-processing units 20-1 to 20-m perform parity update computation, utilizing each element ri,j in R1,3, R2,3 and R3,3 stored in the
memory 30 in the eighth cycle. As a result, Q1,3, Q2,3 and Q3,3 stored in thememory 30 are updated to Q′1,3, Q′2,3 and Q′3,3. - <Tenth Cycle>
- In the tenth cycle, the bit-processing units 10-1 to 10-M perform the same addition process as in the third cycle, using Q′1,3, Q′2,3 and Q′3,3. Further, in the tenth cycle, the bit-processing units 10-1 to 10-M perform a subtraction process in parallel with the addition process. In the subtraction process, the final result of the addition process held by the one of the
register 111 a in each bit-processing unit 10-i (i=1, 2, . . . , M), and Q1,5, Q2,5 and Q3,5 stored in thememory 30 are used. The final result of the addition process means the sum of ψ(qi,j) values corresponding to positions in the ith row of the check matrix H, at which the bit is set to “1”. In the subtraction process, ψ(qi,j5) is subtracted from the final result of the addition process. To this end, themultiplexer 112 b of thesubtraction circuit 112 selects the one of theregisters 111 a that holds the final result of the addition process. Themultiplexer 112 a of thesubtraction circuit 112 selects ψ(qi,j5). After that, the subtracter 112 c of thesubtraction circuit 112 subtracts ψ(qi,j5) selected by themultiplexer 112 a from the final result of the addition process selected by themultiplexer 112 b. - On the other hand, the
sign computation circuit 130 of each of the bit-processing units 10-1 to 10-M excludes MSB(qi,j5) from the final result of the exclusive-OR computation. To this end, themultiplexer 136 in thesign computation circuit 130 selects the one of theregisters 131 that holds the final result of the exclusive-OR computation. Further, themultiplexer 135 in thesign computation circuit 130 selects MSB(qi,j5). After that, the exclusive-OR circuit 137 of thesign computation circuit 130 acquires the exclusive-OR of the final result of the exclusive-OR computation selected by themultiplexer 136 and MSB(qi,j5) selected by themultiplexer 135. In accordance with the computation result of the exclusive-OR circuit 137, the sign of information output from themultiplexer 138 is determined, whereby bit information ri,j is acquired. The bit information ri,j is stored in thememory 30 at a position determined from the ith row and jth column corresponding to the bit information ri,j. In the tenth cycle, R1,5, R2,5 and R3,5 are generated and stored in thememory 30. R1,5, R2,5 and R3,5 represent a set of ri,j values corresponding to the bit positions in the permuted matrices I1,5, I2,5 and I3,5, which correspond to the (4m+1)th to 5mth (Nth) columns (of the fifth column block) included in the check matrix H, and at which the bit is set to “1”. - Thus, in the embodiment, bit update computation corresponding to each of the first to Mth rows of the check matrix H can be executed in 10 (=2×s) cycles.
- Furthermore, in the tenth cycle, the parity-processing units 20-1 to 20-m perform parity update computation, utilizing each element ri,j in R1,4, R2,4 and R3,4 stored in the
memory 30 in the ninth cycle. As a result, Q1,4, Q2,4 and Q3,4 stored in thememory 30 are updated to Q′1,4, Q′2,4 and Q′3,4. - <Eleventh Cycle>
- In the eleventh cycle, the bit-processing units 10-1 to 10-M perform the same addition process as in the fourth cycle, using Q′1,4, Q′2,4 and Q′3,4.
- Further, in the eleventh cycle, the parity-processing units 20-1 to 20-m perform parity update computation using each element ri,j in R1,5, R2,5 and R3,5 stored in the
memory 30 in the tenth cycle. As a result, Q1,5, Q2,5 and Q3,5 stored in thememory 30 are updated to Q′1,5, Q′2,5 and Q′3,5, which is the completion of the first iteration operation including bit update computation and parity update computation. However, bit update computation included in the second iteration operation is already started in the eighth cycle (i.e., in the cycle following the cycle in which parity update computation of the preceding (first) iteration operation is started). - In the twelfth cycle, et seq., the same operations as those performed in the fifth cycle, et seq., are performed.
FIG. 7 orderly shows the above-described schedule for decoding an LDPC code by the LDPC-code decoder. Note that the iteration operation of bit update computation and parity update computation is performed until a hard decision value, which is determined based on r (=3) qj values corresponding to each of all columns k (j=1, 2, . . . , N) of the check matrix H, satisfies the jth parity-check equation. However, if at least one column that does not satisfy the parity-check equation exists even after the iteration operation is performed a predetermined number of times, it is determined that decoding of the LDPC code has failed, whereby the decoding operation is forcibly terminated. - In the above-described embodiment, each parity-processing unit 20-k is configured to simultaneously perform parity update computation concerning all positions (ith row position) in the jth column of the check matrix H, using corresponding ri,j. However, a parity-processing unit that has the same structure as each bit-processing unit 10-i may be employed instead of each parity-processing unit 20-k. Namely, parity information qi,j corresponding to the ith row position in the jth column of the check matrix H-can be sequentially updated by an addition process and subtraction process in two cycles. In this case, it is sufficient if only a single lookup table is provided at the output side of the parity-processing unit. However, the number of cycles necessary for parity update computation is double the number of cycles in the case of using the parity-processing unit 20-k. Further, the start time of bit update computation is delayed by one cycle. In addition, in the above embodiment, the check matrix H is formed of (3×5) permuted matrices. However, the arrangement of the permuted matrices in the check matrix H is not limited to the above. It is sufficient if the check matrix H is formed of (r×s) permuted matrices, assuming that r<s.
- Additional advantages and modifications will readily occur to those skilled in the art. Therefore, the invention in its broader aspects is not limited to the specific details and representative embodiments shown and described herein. Accordingly, various modifications may be made without departing from the spirit or scope of the general inventive concept as defined by the appended claims and their equivalents.
Claims (14)
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2004-193767 | 2004-06-30 | ||
JP2004193767A JP4282558B2 (en) | 2004-06-30 | 2004-06-30 | Low density parity check code decoder and method |
Publications (2)
Publication Number | Publication Date |
---|---|
US20060005105A1 true US20060005105A1 (en) | 2006-01-05 |
US7500168B2 US7500168B2 (en) | 2009-03-03 |
Family
ID=34978836
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/168,329 Expired - Fee Related US7500168B2 (en) | 2004-06-30 | 2005-06-29 | Decoder and decoding method for decoding low-density parity-check codes with parity check matrix |
Country Status (7)
Country | Link |
---|---|
US (1) | US7500168B2 (en) |
EP (1) | EP1612948B1 (en) |
JP (1) | JP4282558B2 (en) |
KR (1) | KR100738864B1 (en) |
CN (1) | CN1716786A (en) |
DE (1) | DE602005010150D1 (en) |
TW (1) | TWI261973B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080187657A1 (en) * | 2006-09-19 | 2008-08-07 | Altan M Cengiz | Methods and apparatus for depositing nanoparticles on a substrate |
WO2009123728A1 (en) * | 2008-03-31 | 2009-10-08 | Sirius Xm Radio Inc. | Efficient, programmable and scalable low density parity check decoder |
US20090319860A1 (en) * | 2008-06-23 | 2009-12-24 | Ramot At Tel Aviv University Ltd. | Overcoming ldpc trapping sets by decoder reset |
US20100070825A1 (en) * | 2005-12-29 | 2010-03-18 | Andrey Gennadievich Efimov | Fast Low-Density Parity-Check Code Encoder |
US8806307B2 (en) | 2008-06-23 | 2014-08-12 | Ramot At Tel Aviv University Ltd. | Interruption criteria for block decoding |
US8819518B2 (en) | 2005-12-01 | 2014-08-26 | Thomson Licensing | Apparatus and method for decoding low density parity check coded signals |
US20160217030A1 (en) * | 2015-01-23 | 2016-07-28 | Dong-Min Shin | Memory system and method of operating the memory system |
Families Citing this family (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1832001A1 (en) * | 2004-12-29 | 2007-09-12 | Intel Corporation | 3-stripes gilbert low density parity-check codes |
JP4492352B2 (en) * | 2005-01-05 | 2010-06-30 | 日本電気株式会社 | Decoding apparatus, method thereof, and program |
US7634710B2 (en) * | 2005-03-25 | 2009-12-15 | Teranetics, Inc. | Efficient decoding |
JP4519694B2 (en) * | 2005-03-29 | 2010-08-04 | 財団法人北九州産業学術推進機構 | LDPC code detection apparatus and LDPC code detection method |
JP2006339799A (en) * | 2005-05-31 | 2006-12-14 | Toshiba Corp | Irregular low density parity check code decoder and decoding method |
US8832326B1 (en) * | 2005-11-01 | 2014-09-09 | Xilinx, Inc. | Circuit and method for ordering data words |
JP4807063B2 (en) | 2005-12-20 | 2011-11-02 | ソニー株式会社 | Decoding device, control method, and program |
US7783951B2 (en) * | 2006-02-09 | 2010-08-24 | NEC Laboratories, America, Inc. | Multiplexed coding for user cooperation |
JP5215537B2 (en) * | 2006-06-28 | 2013-06-19 | 三星電子株式会社 | Information encoding apparatus, information decoding apparatus, information encoding method, and information decoding method |
US7647548B2 (en) * | 2006-07-31 | 2010-01-12 | Agere Systems Inc. | Methods and apparatus for low-density parity check decoding using hardware-sharing and serial sum-product architecture |
US7644339B2 (en) | 2006-10-02 | 2010-01-05 | Broadcom Corporation | Overlapping sub-matrix based LDPC (low density parity check) decoder |
JP4780027B2 (en) * | 2007-04-27 | 2011-09-28 | ソニー株式会社 | Decoding device, control method, and program |
US8418023B2 (en) * | 2007-05-01 | 2013-04-09 | The Texas A&M University System | Low density parity check decoder for irregular LDPC codes |
JP4760769B2 (en) * | 2007-05-15 | 2011-08-31 | 住友電気工業株式会社 | Decoding circuit and partial processing circuit |
WO2009046534A1 (en) * | 2007-10-11 | 2009-04-16 | The Royal Institution For The Advancement Of Learning/Mcgill University | Methods and apparatuses of mathematical processing |
US8161348B2 (en) * | 2008-02-05 | 2012-04-17 | Agere Systems Inc. | Systems and methods for low cost LDPC decoding |
CN101262231B (en) * | 2008-04-25 | 2011-09-28 | 浙江大学 | A decoding method for block low-density check code and reconstruction of multi-mode decoder |
WO2009156935A1 (en) | 2008-06-23 | 2009-12-30 | Ramot At Tel Aviv University Ltd. | Using damping factors to overcome ldpc trapping sets |
US8201068B2 (en) * | 2009-01-06 | 2012-06-12 | Mediatek Inc. | Method for generating parity-check matrix |
JP5316096B2 (en) * | 2009-03-05 | 2013-10-16 | 日本電気株式会社 | Arithmetic circuit with error detection function and error detection method of arithmetic circuit |
JP4799637B2 (en) * | 2009-04-15 | 2011-10-26 | 株式会社モバイルテクノ | Low density parity check code decoding apparatus and method |
JP4660605B2 (en) | 2009-05-28 | 2011-03-30 | 株式会社東芝 | Decoding device, decoding method, and magnetic disk device |
KR20110124659A (en) | 2010-05-11 | 2011-11-17 | 삼성전자주식회사 | Apparatus and method for layered decoding in communication system using low-density parity-check codes |
JP5171997B2 (en) * | 2011-06-20 | 2013-03-27 | パナソニック株式会社 | Decoding method and decoding apparatus |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050229087A1 (en) * | 2004-04-13 | 2005-10-13 | Sunghwan Kim | Decoding apparatus for low-density parity-check codes using sequential decoding, and method thereof |
US7178081B2 (en) * | 2002-12-24 | 2007-02-13 | Electronics And Telecommunications Research Institute | Simplified message-passing decoder for low-density parity-check codes |
US7181676B2 (en) * | 2004-07-19 | 2007-02-20 | Texas Instruments Incorporated | Layered decoding approach for low density parity check (LDPC) codes |
US7178880B2 (en) * | 2004-06-15 | 2007-02-20 | Andersen James H | Wheel centering pin, kit and method |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2799592B1 (en) | 1999-10-12 | 2003-09-26 | Thomson Csf | SIMPLE AND SYSTEMATIC CONSTRUCTION AND CODING METHOD OF LDPC CODES |
US6539367B1 (en) | 2000-05-26 | 2003-03-25 | Agere Systems Inc. | Methods and apparatus for decoding of general codes on probability dependency graphs |
US6633856B2 (en) | 2001-06-15 | 2003-10-14 | Flarion Technologies, Inc. | Methods and apparatus for decoding LDPC codes |
US6938196B2 (en) * | 2001-06-15 | 2005-08-30 | Flarion Technologies, Inc. | Node processors for use in parity check decoders |
US6895547B2 (en) | 2001-07-11 | 2005-05-17 | International Business Machines Corporation | Method and apparatus for low density parity check encoding of data |
US6829308B2 (en) | 2002-07-03 | 2004-12-07 | Hughes Electronics Corporation | Satellite communication system utilizing low density parity check codes |
ATE556491T1 (en) | 2002-07-03 | 2012-05-15 | Dtvg Licensing Inc | METHOD AND PROCEDURE FOR MEMORY MANAGEMENT IN LOW DENSITY PARITY CHECK (LDPC) DECODERING |
US20040019845A1 (en) | 2002-07-26 | 2004-01-29 | Hughes Electronics | Method and system for generating low density parity check codes |
US7222289B2 (en) | 2002-09-30 | 2007-05-22 | Certance Llc | Channel processor using reduced complexity LDPC decoder |
KR100502609B1 (en) * | 2002-11-21 | 2005-07-20 | 한국전자통신연구원 | Encoder using low density parity check code and encoding method thereof |
JP4062435B2 (en) | 2002-12-03 | 2008-03-19 | 日本電気株式会社 | Error correction code decoding apparatus |
KR100996029B1 (en) * | 2003-04-29 | 2010-11-22 | 삼성전자주식회사 | Apparatus and method for coding of low density parity check code |
KR20040101743A (en) * | 2003-05-26 | 2004-12-03 | 삼성전자주식회사 | Apparatus and method for decoding of ldpc in a communication system |
-
2004
- 2004-06-30 JP JP2004193767A patent/JP4282558B2/en not_active Expired - Fee Related
-
2005
- 2005-06-16 TW TW094120025A patent/TWI261973B/en not_active IP Right Cessation
- 2005-06-17 EP EP05105370A patent/EP1612948B1/en not_active Expired - Fee Related
- 2005-06-17 DE DE602005010150T patent/DE602005010150D1/en not_active Expired - Fee Related
- 2005-06-29 US US11/168,329 patent/US7500168B2/en not_active Expired - Fee Related
- 2005-06-30 KR KR1020050057667A patent/KR100738864B1/en not_active IP Right Cessation
- 2005-06-30 CN CNA2005100818425A patent/CN1716786A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7178081B2 (en) * | 2002-12-24 | 2007-02-13 | Electronics And Telecommunications Research Institute | Simplified message-passing decoder for low-density parity-check codes |
US20050229087A1 (en) * | 2004-04-13 | 2005-10-13 | Sunghwan Kim | Decoding apparatus for low-density parity-check codes using sequential decoding, and method thereof |
US7178880B2 (en) * | 2004-06-15 | 2007-02-20 | Andersen James H | Wheel centering pin, kit and method |
US7181676B2 (en) * | 2004-07-19 | 2007-02-20 | Texas Instruments Incorporated | Layered decoding approach for low density parity check (LDPC) codes |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8819518B2 (en) | 2005-12-01 | 2014-08-26 | Thomson Licensing | Apparatus and method for decoding low density parity check coded signals |
US20100070825A1 (en) * | 2005-12-29 | 2010-03-18 | Andrey Gennadievich Efimov | Fast Low-Density Parity-Check Code Encoder |
US8271850B2 (en) * | 2005-12-29 | 2012-09-18 | Intel Corporation | Fast low-density parity-check code encoder |
US20080187657A1 (en) * | 2006-09-19 | 2008-08-07 | Altan M Cengiz | Methods and apparatus for depositing nanoparticles on a substrate |
WO2009123728A1 (en) * | 2008-03-31 | 2009-10-08 | Sirius Xm Radio Inc. | Efficient, programmable and scalable low density parity check decoder |
US20100064199A1 (en) * | 2008-03-31 | 2010-03-11 | Sirius Xm Radio Inc. | Efficient, programmable and scalable low density parity check decoder |
US8601342B2 (en) * | 2008-03-31 | 2013-12-03 | Sirius Xm Radio Inc. | Efficient, programmable and scalable low density parity check decoder |
US9160366B2 (en) | 2008-03-31 | 2015-10-13 | Sirius Xm Radio Inc. | Efficient, programmable and scalable low density parity check decoder |
US20090319860A1 (en) * | 2008-06-23 | 2009-12-24 | Ramot At Tel Aviv University Ltd. | Overcoming ldpc trapping sets by decoder reset |
US8806307B2 (en) | 2008-06-23 | 2014-08-12 | Ramot At Tel Aviv University Ltd. | Interruption criteria for block decoding |
US20160217030A1 (en) * | 2015-01-23 | 2016-07-28 | Dong-Min Shin | Memory system and method of operating the memory system |
US10007572B2 (en) * | 2015-01-23 | 2018-06-26 | Samsung Electronics Co., Ltd. | Memory system and method of operating the memory system |
Also Published As
Publication number | Publication date |
---|---|
JP4282558B2 (en) | 2009-06-24 |
KR20060048750A (en) | 2006-05-18 |
EP1612948A1 (en) | 2006-01-04 |
TWI261973B (en) | 2006-09-11 |
EP1612948B1 (en) | 2008-10-08 |
KR100738864B1 (en) | 2007-07-16 |
US7500168B2 (en) | 2009-03-03 |
CN1716786A (en) | 2006-01-04 |
JP2006019889A (en) | 2006-01-19 |
DE602005010150D1 (en) | 2008-11-20 |
TW200607248A (en) | 2006-02-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7500168B2 (en) | Decoder and decoding method for decoding low-density parity-check codes with parity check matrix | |
US7451376B2 (en) | Decoder and decoding method for decoding irregular low-density parity-check codes | |
US10958288B2 (en) | Decoder for low-density parity-check codes | |
US7730377B2 (en) | Layered decoding of low density parity check (LDPC) codes | |
JP4519694B2 (en) | LDPC code detection apparatus and LDPC code detection method | |
US7127659B2 (en) | Memory efficient LDPC decoding methods and apparatus | |
US9432053B1 (en) | High speed LDPC decoder | |
EP1909395B1 (en) | Communication apparatus and decoding method | |
JP4320418B2 (en) | Decoding device and receiving device | |
TWI699977B (en) | Method employed in ldpc decoder and the decoder | |
US20070089019A1 (en) | Error correction decoder, method and computer program product for block serial pipelined layered decoding of structured low-density parity-check (LDPC) codes, including calculating check-to-variable messages | |
US20040153938A1 (en) | Error correcting code decoding device, program and method used in the same | |
US20070089016A1 (en) | Block serial pipelined layered decoding architecture for structured low-density parity-check (LDPC) codes | |
US20060015802A1 (en) | Layered decoding approach for low density parity check (LDPC) codes | |
Yuan et al. | Successive cancellation list polar decoder using log-likelihood ratios | |
US7434135B2 (en) | Decoder and method for decoding low-density parity-check code | |
US7870468B1 (en) | Reed-solomon decoder using a configurable arithmetic processor | |
TWI551059B (en) | MULTI-CODE CHIEN'S SEARCH CIRCUIT FOR BCH CODES WITH VARIOUS VALUES OF M IN GF(2m) | |
Amaricai et al. | Generalized Very High Throughput Unrolled LDPC Layered Decoder | |
Osman et al. | Two extended programmable BCH soft decoders using least reliable bits reprocessing | |
JP5434454B2 (en) | Decryption device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: KABUSHIKI KAISHA TOSHIBA, JAPAN Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:YOSHIDA, KENJI;REEL/FRAME:016742/0125 Effective date: 20050616 |
|
FEPP | Fee payment procedure |
Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
FPAY | Fee payment |
Year of fee payment: 4 |
|
FPAY | Fee payment |
Year of fee payment: 8 |
|
FEPP | Fee payment procedure |
Free format text: MAINTENANCE FEE REMINDER MAILED (ORIGINAL EVENT CODE: REM.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
LAPS | Lapse for failure to pay maintenance fees |
Free format text: PATENT EXPIRED FOR FAILURE TO PAY MAINTENANCE FEES (ORIGINAL EVENT CODE: EXP.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
STCH | Information on status: patent discontinuation |
Free format text: PATENT EXPIRED DUE TO NONPAYMENT OF MAINTENANCE FEES UNDER 37 CFR 1.362 |
|
FP | Lapsed due to failure to pay maintenance fee |
Effective date: 20210303 |